Daily Papers

Mixture-of-Experts Multimodal Large Language Models (MoE MLLMs) suffer from a significant efficiency bottleneck during Expert Parallelism (EP) inference due to the straggler effect. This issue is worsened in the multimodal context, as existing token-count-based load balancing methods fail to address two unique challenges: (1) Information Heterogeneity, where numerous redundant visual tokens are treated equally to semantically critical ones, and (2) Modality Dynamics, where varying visual to text ratios across tasks lead to resource misallocation. To address these challenges, we propose MACS (Modality-Aware Capacity Scaling), a training-free inference framework. Specifically, MACS introduces an Entropy-Weighted Load mechanism to quantify the semantic value of visual tokens, addressing information heterogeneity. Additionally, the Dynamic Modality-Adaptive Capacity mechanism allocates expert resources based on the real-time modal composition of the input. Extensive experiments demonstrate that MACS significantly outperforms existing methods on various multimodal benchmarks, providing a novel and robust solution for the efficient deployment of MoE MLLMs in EP inference.

In large-scale AI training, Sparse Mixture-of-Experts (s-MoE) layers enable scaling by activating only a small subset of experts per token. An operational challenge in this design is load balancing: routing tokens to minimize the number of idle experts, which is important for the efficient utilization of (costly) GPUs. We provide a theoretical framework for analyzing the Auxiliary-Loss-Free Load Balancing (ALF-LB) procedure -- proposed by DeepSeek's Wang et al. (2024) -- by casting it as a one-step-per-iteration primal-dual method for an assignment problem. First, in a stylized deterministic setting, our framework yields several insightful structural properties: (i) a monotonic improvement of a Lagrangian objective, (ii) a preference rule that moves tokens from overloaded to underloaded experts, and (iii) an approximate-balancing guarantee. Then, we incorporate the stochastic and dynamic nature of AI training using a generalized online optimization formulation. In the online setting, we derive a strong convexity property of the objective that leads to a logarithmic expected regret bound under certain step-size choices. Additionally, we present real experiments on 1B-parameter DeepSeekMoE models to complement our theoretical findings. Together, these results build a principled framework for analyzing the Auxiliary-Loss-Free Load Balancing of s-MoE in AI models.

Efficient parallelization of Large Language Models (LLMs) with long sequences is essential but challenging due to their significant computational and memory demands, particularly stemming from communication bottlenecks in attention mechanisms. While sequence parallelism (SP) has been introduced as a potential solution, existing methods often suffer from limited scalability or inefficiency, rendering their effectiveness. Ring-Attention demonstrates the potential for scaling sequence processing but faces significant limitations due to its reliance on peer-to-peer (P2P) communication and inefficient utilization of network resources. As the degree of SP increases, the quadratic decrease in computation time per step contrasts sharply with the linear reduction in communication volume, exacerbating communication bottlenecks. To address these challenges, we propose TokenRing, a fine-grained parallel framework that leverages bidirectional P2P communication to effectively overlap computation and data transmission. By partitioning the attention block and concurrently transmitting Query and block outputs (i.e., block_out and block_lse) within a fully connected mesh topology, TokenRing achieves significant reductions in communication overhead and better load balancing. These innovations improve the scalability and efficiency of distributed Transformer models, particularly for long-context sequences. Experimental results demonstrate that TokenRing enhances throughput and reduces communication latency. Moreover, its design adapts seamlessly to various multi-GPU interconnect solutions, such as Huawei Ascend, ensuring broad compatibility and cost-effectiveness for distributed LLM inference and training. The code is available at: https://github.com/ACA-Lab-SJTU/token-ring.

Heterogeneous LLM serving stacks split scheduling into two layers that optimize in isolation: model routers pick a model from quality and cost signals while ignoring instance load, and serving load balancers optimize queues while ignoring quality. We present RouteBalance, a serving-aware scheduling layer that fuses both into a single online assignment over concrete model instances, jointly trading off quality, latency, and cost. A batched in-process predictor stack and dead-reckoned instance state keep the joint decision cheap on the request hot path (approx32 ms at 12 req/s). On a 13-instance, 28-GPU heterogeneous cluster serving four model sizes, a single deployed RouteBalance stack traces the upper region of the three-way quality-cost-throughput frontier. Sweeping one weight vector reaches both the highest routing-decision quality (DeepEval 0.419, +0.013 over the strongest baseline, 95% CI [{+}0.005,{+}0.022]; the ordering holds when a second judge re-scores the actually served text) and, at its cost-priority corner, per-request cost that ties the cheapest baseline. With router engineering equalized against concurrent-scoring baseline variants we build, its balanced preset serves at 2.8 s and 30 req/s, leading 2.6 to 4.1times ahead of enhanced BEST-Route at high load. (Deploying those routers as published, one serial scoring call per request, makes them collapse 23times under load, a deployment-architecture effect we isolate separately, not the routing result.) A four-arm isolation shows the benefit follows from pricing latency at model-selection time; the learned predictors contribute calibration and SLO headroom rather than the headline frontier. Code: https://github.com/AKafakA/route-balance

Large language model (LLM) inference workload dominates a wide variety of modern AI applications, ranging from multi-turn conversation to document analysis. Balancing fairness and efficiency is critical for managing diverse client workloads with varying prefix patterns. Unfortunately, existing fair scheduling algorithms for LLM serving, such as Virtual Token Counter (VTC), fail to take prefix locality into consideration and thus suffer from poor performance. On the other hand, locality-aware scheduling algorithms in existing LLM serving frameworks tend to maximize the prefix cache hit rate without considering fair sharing among clients. This paper introduces the first locality-aware fair scheduling algorithm, Deficit Longest Prefix Match (DLPM), which can maintain a high degree of prefix locality with a fairness guarantee. We also introduce a novel algorithm, Double Deficit LPM (D^2LPM), extending DLPM for the distributed setup that can find a balance point among fairness, locality, and load-balancing. Our extensive evaluation demonstrates the superior performance of DLPM and D^2LPM in ensuring fairness while maintaining high throughput (up to 2.87times higher than VTC) and low per-client (up to 7.18times lower than state-of-the-art distributed LLM serving system) latency.

Mixture-of-Experts (MoE) models are typically pre-trained with explicit load-balancing constraints to ensure statistically balanced expert routing. Despite this, we observe that even well-trained MoE models exhibit significantly imbalanced routing. This behavior is arguably natural-and even desirable - as imbalanced routing allows models to concentrate domain-specific knowledge within a subset of experts. Expert parallelism (EP) is designed to scale MoE models by distributing experts across multiple devices, but with a less-discussed assumption of balanced routing. Under extreme imbalance, EP can funnel a disproportionate number of tokens to a small number of experts, leading to compute- and memory-bound failures on overloaded devices during post-training or inference, where explicit load balancing is often inapplicable. We propose Least-Loaded Expert Parallelism (LLEP), a novel EP algorithm that dynamically reroutes excess tokens and associated expert parameters from overloaded devices to underutilized ones. This ensures that all devices complete their workloads within the minimum collective latency while respecting memory constraints. Across different model scales, LLEP achieves up to 5x speedup and 4x reduction in peak memory usage compared to standard EP. This enables faster and higher-throughput post-training and inference, with ~1.9x faster for gpt-oss-120b. We support our method with extensive theoretical analysis and comprehensive empirical evaluations, including ablation studies. These results illuminate key trade-offs and enable a principled framework for hardware-specific hyper-parameter tuning to achieve optimal performance.

This paper presents Block, a distributed scheduling framework designed to optimize load balancing and auto-provisioning across instances in large language model serving frameworks by leveraging contextual information from incoming requests. Unlike popular model serving systems that rely on monolithic and heuristic task schedulers, Block operates as a fully distributed, stateless, and predictive scheduling system to achieve low overhead, reliability, and scalability. It leverages the deterministic and predictable characteristics of LLM inferences, such as host configurations, response lengths, and hardware performance, to make scheduling decisions based on accurately predicted metrics. Evaluation on a 12 GPUs cluster shows that Block significantly outperforms heuristic schedulers, boosting serving capacity by up to 16.7\% and reducing P99 tail latency by up to 49.5\%. These performance gains remain consistent across diverse models, workloads and configurations. Code and data are open-sourced.

Large Language Models (LLMs) typically generate outputs token by token using a fixed compute budget, leading to inefficient resource utilization. To address this shortcoming, recent advancements in mixture of expert (MoE) models, speculative decoding, and early exit strategies leverage the insight that computational demands can vary significantly based on the complexity and nature of the input. However, identifying optimal routing patterns for dynamic execution remains an open challenge, limiting the full potential of these adaptive methods. To address this need, we study adaptive computation in LLMs more systematically. We propose a novel framework that integrates smaller auxiliary modules within each Feed-Forward Network layer of the LLM. This design enables dynamic routing of tokens based on task complexity: tokens can be processed by either the small or big modules at each layer, or even bypass certain layers entirely. This allows us to introduce a novel notion of a token's difficulty, defined by its potential to benefit from additional computational resources. Importantly, by employing oracles to identify optimal patterns of adaptive computations, we gain valuable insights into the internal workings of LLMs and the routing processes in a simplified heterogeneous MoE setup. We show that trained routers operate differently from oracles and often yield suboptimal solutions. Notably, activating a large module in just one layer outperforms models that use large modules across all layers, underscoring the gap between practical implementations of routing in MoE models and theoretical optima for adaptive computation.

For Mixture-of-Experts (MoE) models, an unbalanced expert load will lead to routing collapse or increased computational overhead. Existing methods commonly employ an auxiliary loss to encourage load balance, but a large auxiliary loss will introduce non-negligible interference gradients into training and thus impair the model performance. In order to control load balance while not producing undesired gradients during training, we propose Loss-Free Balancing, featured by an auxiliary-loss-free load balancing strategy. To be specific, before the top-K routing decision, Loss-Free Balancing will first apply an expert-wise bias to the routing scores of each expert. By dynamically updating the bias of each expert according to its recent load, Loss-Free Balancing can consistently maintain a balanced distribution of expert load. In addition, since Loss-Free Balancing does not produce any interference gradients, it also elevates the upper bound of model performance gained from MoE training. We validate the performance of Loss-Free Balancing on MoE models with up to 3B parameters trained on up to 200B tokens. Experimental results show that Loss-Free Balancing achieves both better performance and better load balance compared with traditional auxiliary-loss-controlled load balancing strategies.

Serving numerous users and requests concurrently requires good fairness in Large Language Models (LLMs) serving system. This ensures that, at the same cost, the system can meet the Service Level Objectives (SLOs) of more users , such as time to first token (TTFT) and time between tokens (TBT), rather than allowing a few users to experience performance far exceeding the SLOs. To achieve better fairness, the preemption-based scheduling policy dynamically adjusts the priority of each request to maintain balance during runtime. However, existing systems tend to overly prioritize throughput, overlooking the overhead caused by preemption-induced context switching, which is crucial for maintaining fairness through priority adjustments. In this work, we identify three main challenges that result in this overhead. 1) Inadequate I/O utilization. 2) GPU idleness. 3) Unnecessary I/O transmission during multi-turn conversations. Our key insight is that the block-based KV cache memory policy in existing systems, while achieving near-zero memory waste, leads to discontinuity and insufficient granularity in the KV cache memory. To respond, we introduce FastSwitch, a fairness-aware serving system that not only aligns with existing KV cache memory allocation policy but also mitigates context switching overhead. Our evaluation shows that FastSwitch outperforms the state-of-the-art LLM serving system vLLM with speedups of 1.4-11.2x across different tail TTFT and TBT.

The Mixture of Experts (MoE) is an effective architecture for scaling large language models by leveraging sparse expert activation, optimizing the trade-off between performance and efficiency. However, under expert parallelism, MoE suffers from inference inefficiencies due to imbalanced token-to-expert assignment, where some experts are overloaded while others remain underutilized. This imbalance leads to poor resource utilization and increased latency, as the most burdened expert dictates the overall delay, a phenomenon we define as the \textit{Straggler Effect}. To mitigate this, we propose Capacity-Aware Inference, including two key techniques: (1) \textit{Capacity-Aware Token Drop}, which discards overloaded tokens to regulate the maximum latency of MoE, and (2) \textit{Capacity-Aware Token Reroute}, which reallocates overflowed tokens to underutilized experts, balancing the token distribution. These techniques collectively optimize both high-load and low-load expert utilization, leading to a more efficient MoE inference pipeline. Extensive experiments demonstrate the effectiveness of our methods, showing significant improvements in inference efficiency, e.g., 0.2\% average performance increase and a 1.94times inference speedup on Mixtral-8times7B-Instruct.

We model Mixture-of-Experts (MoE) token routing as a congestion game with a single effective parameter, the congestion coefficient gamma_eff, that quantifies the balance-quality tradeoff. Tracking gamma_eff across training checkpoints of two open-source MoE models, OLMoE-1B-7B (20 checkpoints, with dense sampling in the surge region) and OpenMoE-8B (6 checkpoints), reveals a three-phase trajectory: a surge phase where the router learns to balance load (gamma_eff: 14 to 36-39, peaking in the step 30K-40K region), a stabilization phase where experts specialize under steady balance (B_0: 2.4 to 2.3, steps 100K-400K), and a relaxation phase where the router trades balance for quality as experts differentiate (gamma_eff: 27 to 9, steps 400K-1.2M). This non-monotone trajectory, invisible to post-hoc analysis of converged models, reveals that early MoE training prioritizes balance while late training prioritizes quality. The theoretical framework is honest about its limits: the single-type equilibrium reduces to temperature-scaled softmax (held-out L1: MFG = 0.199 vs. softmax = 0.200). The game is not a better predictor; it reveals what the temperature means and, critically, how that temperature evolves. We complement the dynamics with an effective congestion decomposition, a multi-type extension that improves load prediction via token clustering on all 16 layers (mean: 30%), scope diagnostics (K/M, epsilon_l), and robustness verification across four independent quality estimators (r >= 0.89). All confidence intervals are from bootstrap resampling over 50 independent text batches.

The scaling of Large Language Models (LLMs) drives interest in matrix-based optimizers (e.g., Shampoo, Muon, SOAP) for their convergence efficiency; yet their requirement for holistic updates conflicts with the tensor fragmentation in distributed frameworks like Megatron. Existing solutions are suboptimal: synchronous approaches suffer from computational redundancy, while layer-wise partitioning fails to reconcile this conflict without violating the geometric constraints of efficient communication primitives. To bridge this gap, we propose Canzona, a Unified, Asynchronous, and Load-Balanced framework that decouples logical optimizer assignment from physical parameter distribution. For Data Parallelism, we introduce an alpha-Balanced Static Partitioning strategy that respects atomicity while neutralizing the load imbalance. For Tensor Parallelism, we design an Asynchronous Compute pipeline utilizing Micro-Group Scheduling to batch fragmented updates and hide reconstruction overhead. Extensive evaluations on the Qwen3 model family (up to 32B parameters) on 256 GPUs demonstrate that our approach preserves the efficiency of established parallel architectures, achieving a 1.57x speedup in end-to-end iteration time and reducing optimizer step latency by 5.8x compared to the baseline.

Acceleration of graph applications on GPUs has found large interest due to the ubiquitous use of graph processing in various domains. The inherent irregularity in graph applications leads to several challenges for parallelization. A key challenge, which we address in this paper, is that of load-imbalance. If the work-assignment to threads uses node-based graph partitioning, it can result in skewed task-distribution, leading to poor load-balance. In contrast, if the work-assignment uses edge-based graph partitioning, the load-balancing is better, but the memory requirement is relatively higher. This makes it unsuitable for large graphs. In this work, we propose three techniques for improved load-balancing of graph applications on GPUs. Each technique brings in unique advantages, and a user may have to employ a specific technique based on the requirement. Using Breadth First Search and Single Source Shortest Paths as our processing kernels, we illustrate the effectiveness of each of the proposed techniques in comparison to the existing node-based and edge-based mechanisms.

Diffusion language models (DLMs) enable parallel, non-autoregressive text generation, yet existing DLM mixture-of-experts (MoE) models inherit token-choice (TC) routing from autoregressive systems, leading to load imbalance and rigid computation allocation. We show that expert-choice (EC) routing is a better fit for DLMs: it provides deterministic load balancing by design, yielding higher throughput and faster convergence than TC. Building on the property that EC capacity is externally controllable, we introduce timestep-dependent expert capacity, which varies expert allocation according to the denoising step. We find that allocating more capacity to low-mask-ratio steps consistently achieves the best performance under matched FLOPs, and provide a mechanistic explanation: tokens in low-mask-ratio contexts exhibit an order-of-magnitude higher learning efficiency, so concentrating compute on these steps yields the largest marginal return. Finally, we show that existing pretrained TC DLMs can be retrofitted to EC by replacing only the router, achieving faster convergence and improved accuracy across diverse downstream tasks. Together, these results establish EC routing as a superior paradigm for DLM MoE models and demonstrate that computation in DLMs can be treated as an adaptive policy rather than a fixed architectural constant. Code is available at https://github.com/zhangshuibai/EC-DLM.

This report presents the Prime Collective Communications Library (PCCL), a novel fault-tolerant collective communication library designed for distributed ML workloads over the public internet. PCCL introduces a new programming model that enables dynamic peer joining and failure recovery. The library implements efficient collective operations like all-reduce while providing robust fault tolerance mechanisms that allow the system to continue operating even when peers fail or join during ongoing operations. We demonstrate that PCCL's design enables practical solutions to dynamic membership challenges in workloads with repeated operations and deterministic state advancement. Our implementation passes extensive stress tests across all major operating systems, showing reliable operation even under rapid peer churn and concurrent collective operations. By dispatching to multiple connections, we can efficiently utilize cross-continental long-fat-pipe TCP WAN links, in our experiments achieving up to 45 Gbit/s of bandwidth utilization across Europe and 25 Gbit/s across North America and Europe. PCCL's architecture enables easy implementation of distributed low-communication optimization strategies like DiLoCo, which significantly reduce communication frequency. Combined with quantization, this leads to a significant reduction in the bandwidth required for distributed training workloads. PCCL also allows for concurrent collective operations, which enables optimization strategies like async DiLoCo, which can completely hide communication overhead by implementing one-step delayed parameter updates. PCCL can facilitate exact bit-parity of the shared state across peers in all cases induced by graceful or abrupt peer churn. While PCCL exposes a C99 API, Python bindings are available which are compatible with PyTorch alongside FSDP. PCCL is available under the open source MIT license.

Multi-agent LLM applications organize execution in synchronized rounds where a central scheduler gathers outputs from all agents and redistributes the combined context. This All-Gather communication pattern creates massive KV Cache redundancy, because every agent's prompt contains the same shared output blocks, yet existing reuse methods fail to exploit it efficiently. We present TokenDance, a system that scales the number of concurrent agents by exploiting the All-Gather pattern for collective KV Cache sharing. TokenDance's KV Collector performs KV Cache reuse over the full round in one collective step, so the cost of reusing a shared block is paid once regardless of agent count. Its Diff-Aware Storage encodes sibling caches as block-sparse diffs against a single master copy, achieving 11-17x compression on representative workloads. Evaluation on GenerativeAgents and AgentSociety shows that TokenDance supports up to 2.7x more concurrent agents than vLLM with prefix caching under SLO requirement, reduces per-agent KV Cache storage by up to 17.5x, and achieves up to 1.9x prefill speedup over per-request position-independent caching.

Token-choice Mixture-of-Experts (TC-MoE) routes each token to a fixed number of experts, limiting dynamic computation allocation and requiring auxiliary losses to maintain load balance. We propose Expert Threshold (ET) routing, where each expert maintains an exponential moving average (EMA) threshold estimated from the global token distribution. At both training and inference, each token is independently routed to an expert if its score exceeds the expert's threshold, enabling dynamic computation allocation while achieving load balance without auxiliary losses. This fully causal mechanism eliminates dependence on other tokens in the batch, making it well-suited for autoregressive language modeling. In pretraining experiments scaling to 2.4B parameters on FineWeb-Edu, ET achieves 0.067 lower cross-entropy loss than TC-MoE, equivalent to reaching the same performance with 1.6times fewer tokens.

The Mixtures-of-Experts (MoE) model is a widespread distributed and integrated learning method for large language models (LLM), which is favored due to its ability to sparsify and expand models efficiently. However, the performance of MoE is limited by load imbalance and high latency of All-To-All communication, along with relatively redundant computation owing to large expert capacity. Load imbalance may result from existing routing policies that consistently tend to select certain experts. The frequent inter-node communication in the All-To-All procedure also significantly prolongs the training time. To alleviate the above performance problems, we propose a novel routing strategy that combines load balance and locality by converting partial inter-node communication to that of intra-node. Notably, we elucidate that there is a minimum threshold for expert capacity, calculated through the maximal angular deviation between the gating weights of the experts and the assigned tokens. We port these modifications on the PanGu-Sigma model based on the MindSpore framework with multi-level routing and conduct experiments on Ascend clusters. The experiment results demonstrate that the proposed LocMoE reduces training time per epoch by 12.68% to 22.24% compared to classical routers, such as hash router and switch router, without impacting the model accuracy.

In LLM serving, reusing the KV cache of prompts across requests is critical for reducing TTFT and serving costs. Cache-affinity scheduling, which co-locates requests with the same prompt prefix to maximize KV cache reuse, often conflicts with load-balancing scheduling that distributes requests evenly across compute instances. Existing schedulers fail to reconcile this trade-off as they operate within a single mapping space, typically applying cache-affinity routing to a subset of requests and load-balanced routing to the rest, without a unified solution to achieve both goals. To address this limitation, we propose DualMap, a dual-mapping scheduling strategy for distributed LLM serving that achieves both cache affinity and load balancing. Its key idea is to map each request to two candidate instances via two independent hash functions based on the request prompt, then intelligently select the better candidate based on current system states. This design increases the likelihood that requests with shared prefixes are co-located, while evenly dispersing distinct prefixes across the cluster via ``the power of two choices''. To make DualMap robust under dynamic and skewed real-world workloads, we incorporate three techniques: 1) SLO-aware request routing, which prioritizes cache affinity but switches to load-aware scheduling when TTFT exceeds the SLO, enhancing load balance without sacrificing cache reuse; 2) hotspot-aware rebalancing, which dynamically migrates requests from overloaded to underloaded instances, mitigating hotspots and rebalancing the system; 3) lightweight dual-hash-ring scaling, which leverages a dual-hash-ring mapping to support fast and low-overhead instance scaling without costly global remapping. Experiments on real-world workloads show that DualMap improves effective request capacity by up to 2.25times under the same TTFT SLO constraints compared with SOTA work.

Modern data parallel (DP) training favors collective communication over parameter servers (PS) for its simplicity and efficiency under balanced workloads. However, the balanced workload assumption no longer holds in large language model (LLM) post-training due to the high variance in sequence lengths. Under imbalanced workloads, collective communication creates synchronization barriers, leading to under-utilization of devices with smaller workloads. This change in training dynamics calls for a revisit of the PS paradigm for its robustness to such imbalance. We propose On-Demand Communication (ODC), which adapts PS into Fully Sharded Data Parallel (FSDP) by replacing collective all-gather and reduce-scatter with direct point-to-point communication. Compared to FSDP, ODC reduces the synchronization barrier from once per layer to once per minibatch and decouples the workload on each device so that faster workers are not stalled. It also enables simpler and more effective load balancing at the minibatch level. Across diverse LLM post-training tasks, ODC consistently improves device utilization and training throughput, achieving up to a 36\% speedup over standard FSDP. These results demonstrate that ODC is a superior fit for the prevalent imbalanced workloads in LLM post-training. Our implementation of ODC and integration with FSDP is open-sourced at https://github.com/sail-sg/odc.

We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/

Sparse Mixture-of-Experts (SMoE) has gained prominence as a scalable and computationally efficient architecture, enabling significant growth in model capacity without incurring additional inference costs. However, existing SMoE models often rely on auxiliary losses (e.g., z-loss, load balancing) and additional trainable parameters (e.g., noisy gating) to encourage expert diversity, leading to objective misalignment and increased model complexity. Moreover, existing Sinkhorn-based methods suffer from significant training overhead due to their heavy reliance on the computationally expensive Sinkhorn algorithm. In this work, we formulate token-to-expert assignment as an optimal transport problem, incorporating constraints to ensure balanced expert utilization. We demonstrate that introducing a minimal degree of optimal transport-based routing enhances SMoE performance without requiring auxiliary balancing losses. Unlike previous methods, our approach derives gating scores directly from the transport map, enabling more effective token-to-expert balancing, supported by both theoretical analysis and empirical results. Building on these insights, we propose Selective Sinkhorn Routing (SSR), a routing mechanism that replaces auxiliary loss with lightweight Sinkhorn-based routing. SSR promotes balanced token assignments while preserving flexibility in expert selection. Across both language modeling and image classification tasks, SSR achieves faster training, higher accuracy, and greater robustness to input corruption.

For pre-training of MoE (Mixture-of-Experts) models, one of the main issues is unbalanced expert loads, which may cause routing collapse or increased computational overhead. Existing methods contain the Loss-Controlled method and the Loss-Free method, where both the unbalanced degrees at first several training steps are still high and decrease slowly. In this work, we propose BIP-Based Balancing, an expert load balancing algorithm based on binary integer programming (BIP). The algorithm maintains an additional vector q on each MoE layer that can help change the top-K order of s by solving a binary integer programming with very small time costs. We implement the algorithm on two MoE language models: 16-expert (0.3B) and 64-expert (1.1B). The experimental results show that on both models comparing with the Loss-Controlled method and the Loss-Free method, our algorithm trains models with the lowest perplexities, while saves at least 13% of pre-training time compared with the Loss-Controlled method. Within our current knowledge, this is the first routing algorithm that achieves maintaining load balance status on every expert in every MoE layer from the first step to the last step during the whole pre-training process, while the trained MoE models also perform well. The code material of this work is available at https://github.com/sunyuanLLM/bip_routing_algorithm.

The performance of a language model has been shown to be effectively modeled as a power-law in its parameter count. Here we study the scaling behaviors of Routing Networks: architectures that conditionally use only a subset of their parameters while processing an input. For these models, parameter count and computational requirement form two independent axes along which an increase leads to better performance. In this work we derive and justify scaling laws defined on these two variables which generalize those known for standard language models and describe the performance of a wide range of routing architectures trained via three different techniques. Afterwards we provide two applications of these laws: first deriving an Effective Parameter Count along which all models scale at the same rate, and then using the scaling coefficients to give a quantitative comparison of the three routing techniques considered. Our analysis derives from an extensive evaluation of Routing Networks across five orders of magnitude of size, including models with hundreds of experts and hundreds of billions of parameters.

Training large language models (LLMs) with increasingly long and varying sequence lengths introduces severe load imbalance challenges in large-scale data-parallel training. Recent frameworks attempt to mitigate these issues through data reorganization or hybrid parallel strategies. However, they often overlook how computational and communication costs scale with sequence length, resulting in suboptimal performance. We identify three critical challenges: (1) varying computation-to-communication ratios across sequences of different lengths in distributed attention, (2) mismatch between static NIC-GPU affinity and dynamic parallel workloads, and (3) distinct optimal partitioning strategies required for quadratic attention versus linear components. To address these challenges, we present Zeppelin, a novel training system that integrates three key techniques: (1) a hierarchical sequence partitioning method for the attention module that reduces communication overhead and balances computation, supported by an efficient attention engine that applies divergent parallel strategies; (2) a routing layer that orchestrates inter-node transfers to fully utilize NIC bandwidth; and (3) a remapping layer that transforms sequence layouts between attention and linear modules, ensuring high computational efficiency across both. Comprehensive evaluations across diverse configurations show that Zeppelin delivers an average 2.80x speedup over state-of-the-art methods.

The growing demand for large language models (LLMs) requires serving systems to handle many concurrent requests with diverse service level objectives (SLOs). This exacerbates head-of-line (HoL) blocking during the compute-intensive prefill phase, where long-running requests monopolize resources and delay higher-priority ones, leading to widespread time-to-first-token (TTFT) SLO violations. While chunked prefill enables interruptibility, it introduces an inherent trade-off between responsiveness and throughput: reducing chunk size improves response latency but degrades computational efficiency, whereas increasing chunk size maximizes throughput but exacerbates blocking. This necessitates an adaptive preemption mechanism. However, dynamically balancing execution granularity against scheduling overheads remains a key challenge. In this paper, we propose FlowPrefill, a TTFT-goodput-optimized serving system that resolves this conflict by decoupling preemption granularity from scheduling frequency. To achieve adaptive prefill scheduling, FlowPrefill introduces two key innovations: 1) Operator-Level Preemption, which leverages operator boundaries to enable fine-grained execution interruption without the efficiency loss associated with fixed small chunking; and 2) Event-Driven Scheduling, which triggers scheduling decisions only upon request arrival or completion events, thereby supporting efficient preemption responsiveness while minimizing control-plane overhead. Evaluation on real-world production traces shows that FlowPrefill improves maximum goodput by up to 5.6times compared to state-of-the-art systems while satisfying heterogeneous SLOs.

Large language models (LLMs) have achieved impressive success, but their high memory requirements present challenges for long-context token generation. In this paper we study the streaming complexity of attention approximation, a key computational primitive underlying token generation. Our main contribution is BalanceKV, a streaming algorithm for epsilon-approximating attention computations based on geometric process for selecting a balanced collection of Key and Value tokens as per Banaszczyk's vector balancing theory. We complement our algorithm with space lower bounds for streaming attention computation. Besides strong theoretical guarantees, BalanceKV exhibits empirically validated performance improvements over existing methods, both for attention approximation and end-to-end performance on various long context benchmarks.

Distributed inference of large language models (LLMs) can introduce overheads of up to 20% even over GPUs connected via high-speed interconnects such as NVLINK. Multiple techniques have been proposed to mitigate these overheads by decomposing computations into finer-grained tasks and overlapping communication with sub-tasks as they complete. However, fine-grained decomposition of a large computation into many smaller computations on GPUs results in overheads. Further, the communication itself uses many streaming multiprocessors (SMs), adding to the overhead. We present TokenWeave to address these challenges. TokenWeave proposes a Token-Splitting technique that divides the tokens in the inference batch into two approximately equal subsets in a wave-aware manner. The computation of one subset is then overlapped with the communication of the other. In addition, TokenWeave optimizes the order of the layer normalization computation with respect to communication operations and implements a novel fused AllReduce-RMSNorm kernel carefully leveraging Multimem instruction support available on NVIDIA Hopper GPUs. These optimizations allow TokenWeave to perform communication and RMSNorm using only 2-8 SMs. Moreover, our kernel enables the memory bound RMSNorm to be overlapped with the other batch's computation, providing additional gains. Our evaluations demonstrate up to 29% latency gains and up to 26% throughput gains across multiple models and workloads. In several settings, TokenWeave results in better performance compared to an equivalent model with all communication removed.

Apache Spark is a widely adopted framework for large-scale data processing. However, in industrial analytics environments, Spark's built-in schedulers, such as FIFO and fair scheduling, struggle to maintain both user-level fairness and low mean response time, particularly in long-running shared applications. Existing solutions typically focus on job-level fairness which unintentionally favors users who submit more jobs. Although Spark offers a built-in fair scheduler, it lacks adaptability to dynamic user workloads and may degrade overall job performance. We present the User Weighted Fair Queuing (UWFQ) scheduler, designed to minimize job response times while ensuring equitable resource distribution across users and their respective jobs. UWFQ simulates a virtual fair queuing system and schedules jobs based on their estimated finish times under a bounded fairness model. To further address task skew and reduce priority inversions, which are common in Spark workloads, we introduce runtime partitioning, a method that dynamically refines task granularity based on expected runtime. We implement UWFQ within the Spark framework and evaluate its performance using multi-user synthetic workloads and Google cluster traces. We show that UWFQ reduces the average response time of small jobs by up to 74% compared to existing built-in Spark schedulers and to state-of-the-art fair scheduling algorithms.

Mixture-of-Experts (MoE) models challenge serving infrastructures with dynamic, sparse expert utilization, causing instability on conventional systems designed for dense architectures. We propose EaaS, a novel serving system to enable efficient, scalable, and robust MoE deployment. Our system disaggregates MoE modules into independent, stateless services. This design enables fine-grained resource scaling and provides inherent fault tolerance by decoupling compute units. The architecture is powered by a high-performance, CPU-free peer-to-peer communication library that ensures minimal overhead and high throughput. Experiments confirm EaaS's scalability and efficiency, achieving performance comparable to monolithic systems while providing robust fault tolerance and strong scalability. EaaS incurs less than a 2% throughput reduction under simulated hardware failures that would otherwise halt monolithic architectures. It further saves up to 37.5% of computing resources through dynamic fine-grained adaptation to serving traffic, demonstrating strong resilience for large-scale MoE deployment in production.

This paper introduces Dodoor, an efficient randomized decentralized scheduler designed for task scheduling in modern data centers. Dodoor leverages advanced research on the weighted balls-into-bins model with b-batched setting. Unlike other decentralized schedulers that rely on real-time probing of remote servers, Dodoor makes scheduling decisions based on cached server information, which is updated in batches, to reduce communication overheads. To schedule tasks with dynamic, multidimensional resource requirements in heterogeneous cluster, Dodoor uses a novel load score to measure servers' loads for each scheduled task. This score captures the anti-affinity between servers and tasks in contrast to the commonly used heuristic of counting pending tasks to balance load. On a 101-node heterogeneous cluster, Dodoor is evaluated using two workloads: (i) simulated Azure virtual machines placements and (ii) real serverless Python functions executions in Docker. The evaluation shows that Dodoor reduces scheduling messages by 55--66% on both workloads. Dodoor can also increase throughput by up to 33.2% and 21.5%, reduce mean makespan latency by 12.1% and 7.2%, and improve tail latency by 21.9% and 24.6% across the two workloads.

Large Language Models (LLMs) present a critical trade-off between inference quality and computational cost: larger models offer superior capabilities but incur significant latency, while smaller models are faster but less powerful. Existing serving strategies often employ fixed model scales or static two-stage speculative decoding, failing to dynamically adapt to the varying complexities of user requests or fluctuations in system performance. This paper introduces , a novel framework that reimagines LLM inference as an adaptive routing problem solved through multi-level speculative decoding. dynamically constructs and optimizes inference "paths" (chains of models) based on real-time feedback, addressing the limitations of static approaches. Our contributions are threefold: (1) An adaptive model chain scheduling mechanism that leverages performance profiling (execution times) and predictive similarity metrics (derived from token distribution divergence) to continuously select the optimal sequence of draft and verifier models, minimizing predicted latency per generated token. (2) A multi-level collaborative verification framework where intermediate models within the selected chain can validate speculative tokens, reducing the verification burden on the final, most powerful target model. (3) A synchronized state management system providing efficient, consistent KV cache handling across heterogeneous models in the chain, including precise, low-overhead rollbacks tailored for asynchronous batch processing inherent in multi-level speculation. Preliminary experiments demonstrate the validity of our method.

Mixture-of-Experts (MoE) models have gained popularity in achieving state-of-the-art performance in a wide range of tasks in computer vision and natural language processing. They effectively expand the model capacity while incurring a minimal increase in computation cost during training. However, deploying such models for inference is difficult due to their large size and complex communication pattern. In this work, we provide a characterization of two MoE workloads, namely Language Modeling (LM) and Machine Translation (MT) and identify their sources of inefficiencies at deployment. We propose three optimization techniques to mitigate sources of inefficiencies, namely (1) Dynamic gating, (2) Expert Buffering, and (3) Expert load balancing. We show that dynamic gating improves maximum throughput by 6.21-11.23times for LM, 5.75-10.98times for MT Encoder and 2.58-5.71times for MT Decoder. It also reduces memory usage by up to 1.36times for LM and up to 1.1times for MT. We further propose Expert Buffering, a new caching mechanism that only keeps hot, active experts in GPU memory while buffering the rest in CPU memory. This reduces static memory allocation by up to 1.47times. We finally propose a load balancing methodology that provides additional scalability to the workload.

The integration of Large Language Models (LLMs) into applications ranging from interactive chatbots to multi-agent systems has introduced a wide spectrum of service-level objectives (SLOs) for responsiveness. These include latency-sensitive requests emphasizing per-token latency in streaming chat, deadline-sensitive requests requiring rapid full responses to trigger external tools, and compound requests with evolving dependencies across multiple LLM calls. Despite-or perhaps, because of-this workload diversity and unpredictable request information (e.g., response lengths and dependencies), existing request schedulers have focused on aggregate performance, unable to ensure application-level SLO needs. This paper presents JITServe, the first SLO-aware LLM serving system designed to maximize service goodput (e.g., the number of tokens meeting request SLOs) across diverse workloads. JITServe novelly schedules requests using imprecise request information and gradually relaxes this conservatism by refining request information estimates as generation progresses. It applies a grouped margin goodput maximization algorithm to allocate just enough serving bandwidth to satisfy each request's SLO just-in-time (JIT), maximizing residual capacity for others, while deciding the composition of requests in a batch to maximize efficiency and goodput with provable guarantees. Our evaluation across diverse realistic workloads, including chat, deep research, and agentic pipelines, shows that JITServe improves service goodput by 1.4x-6.3x, alternatively achieving 28.5%-83.2% resource savings, compared to state-of-the-art designs.

Federated Learning (FL) typically involves a large-scale, distributed system with individual user devices/servers training models locally and then aggregating their model updates on a trusted central server. Existing systems for FL often use an always-on server for model aggregation, which can be inefficient in terms of resource utilization. They may also be inelastic in their resource management. This is particularly exacerbated when aggregating model updates at scale in a highly dynamic environment with varying numbers of heterogeneous user devices/servers. We present LIFL, a lightweight and elastic serverless cloud platform with fine-grained resource management for efficient FL aggregation at scale. LIFL is enhanced by a streamlined, event-driven serverless design that eliminates the individual heavy-weight message broker and replaces inefficient container-based sidecars with lightweight eBPF-based proxies. We leverage shared memory processing to achieve high-performance communication for hierarchical aggregation, which is commonly adopted to speed up FL aggregation at scale. We further introduce locality-aware placement in LIFL to maximize the benefits of shared memory processing. LIFL precisely scales and carefully reuses the resources for hierarchical aggregation to achieve the highest degree of parallelism while minimizing the aggregation time and resource consumption. Our experimental results show that LIFL achieves significant improvement in resource efficiency and aggregation speed for supporting FL at scale, compared to existing serverful and serverless FL systems.

Large Language Model (LLM) inference in production must meet stringent service-level objectives for both time-to-first-token (TTFT) and time-between-token (TBT) while maximizing throughput under fixed compute, memory, and interconnect budgets. Modern serving systems adopt stall-free scheduling techniques such as chunked prefill, which splits long prompt processing along the token dimension and interleaves prefill with ongoing decode iterations. While effective at stabilizing TBT, chunked prefill incurs substantial overhead in Mixture-of-Experts (MoE) models: redundant expert weight loads increase memory traffic by up to 39% and inflate energy consumption. We propose layered prefill, a new scheduling paradigm that treats transformer layer groups as the primary scheduling unit. By vertically partitioning the model into contiguous layer groups and interleaving prefill and decode across the groups, layered prefill sustains stall-free decoding while eliminating chunk-induced MoE weight reloads. It reduces off-chip bandwidth demand, lowering TTFT by up to 70%, End-to-End latency by 41% and per-token energy by up to 22%. Evaluations show that layered prefill consistently improves the TTFT--TBT Pareto frontier over chunked prefill, reducing expert-load traffic and energy cost while maintaining stall-free decoding. Overall, shifting the scheduling axis from tokens to layers unlocks a new operating regime for high-efficiency, energy-aware LLM serving in co-located environments.

Large language models (LLMs) are useful in many NLP tasks and become more capable with size, with the best open-source models having over 50 billion parameters. However, using these 50B+ models requires high-end hardware, making them inaccessible to most researchers. In this work, we investigate methods for cost-efficient inference and fine-tuning of LLMs, comparing local and distributed strategies. We observe that a large enough model (50B+) can run efficiently even on geodistributed devices in a consumer-grade network. This could allow running LLM efficiently by pooling together idle compute resources of multiple research groups and volunteers. We address two open problems: (1) how to perform inference and fine-tuning reliably if any device can disconnect abruptly and (2) how to partition LLMs between devices with uneven hardware, joining and leaving at will. In order to do that, we develop special fault-tolerant inference algorithms and load-balancing protocols that automatically assign devices to maximize the total system throughput. We showcase these algorithms in Petals - a decentralized system that runs Llama 2 (70B) and BLOOM (176B) over the Internet up to 10x faster than offloading for interactive generation. We evaluate the performance of our system in simulated conditions and a real-world setup spanning two continents.

Mixture of experts (MoE) models achieve state-of-the-art results in language modeling but suffer from inefficient hardware utilization due to imbalanced token routing and communication overhead. While prior work has focused on optimizing MoE training and decoder architectures, inference for encoder-based MoE models in a multi-GPU with expert parallelism setting remains underexplored. We introduce MoEShard, an inference system that achieves perfect load balancing through tensor sharding of MoE experts. Unlike existing approaches that rely on heuristic capacity factors or drop tokens, MoEShard evenly distributes computation across GPUs and ensures full token retention, maximizing utilization regardless of routing skewness. We achieve this through a strategic row- and column-wise decomposition of expert matrices. This reduces idle time and avoids bottlenecks caused by imbalanced expert assignments. Furthermore, MoEShard minimizes kernel launches by fusing decomposed expert computations, significantly improving throughput. We evaluate MoEShard against DeepSpeed on encoder-based architectures, demonstrating speedups of up to 6.4times in time to first token (TTFT). Our results show that tensor sharding, when properly applied to experts, is a viable and effective strategy for efficient MoE inference.

Efficient inference for on-device Large Language Models (LLMs) remains challenging due to limited hardware resources and the high cost of the prefill stage, which processes the full input context to construct Key-Value (KV) caches. We present SparKV, an adaptive KV loading framework that combines cloud-based KV streaming with on-device computation. SparKV models the cost of individual KV chunks and decides whether each chunk should be streamed or computed locally, while overlapping the two execution paths to reduce latency. To handle fluctuations in wireless connectivity and edge resource availability, SparKV further refines offline-generated schedules at runtime to rebalance communication and computation costs. Experiments across diverse datasets, LLMs, and edge devices show that SparKV reduces Time-to-First-Token by 1.3$x-5.1x with negligible impact on response quality, while lowering per-request energy consumption by 1.5x to 3.3x, demonstrating its robustness and practicality for real-world on-device deployment.

The rapid evolution of large language models (LLMs), driven by growing parameter scales, adoption of mixture-of-experts (MoE) architectures, and expanding context lengths, imposes unprecedented demands on AI infrastructure. Traditional AI clusters face limitations in compute intensity, memory bandwidth, inter-chip communication, and latency, compounded by variable workloads and strict service-level objectives. Addressing these issues requires fundamentally redesigned hardware-software integration. This paper introduces Huawei CloudMatrix, a next-generation AI datacenter architecture, realized in the production-grade CloudMatrix384 supernode. It integrates 384 Ascend 910C NPUs and 192 Kunpeng CPUs interconnected via an ultra-high-bandwidth Unified Bus (UB) network, enabling direct all-to-all communication and dynamic pooling of resources. These features optimize performance for communication-intensive operations, such as large-scale MoE expert parallelism and distributed key-value cache access. To fully leverage CloudMatrix384, we propose CloudMatrix-Infer, an advanced LLM serving solution incorporating three core innovations: a peer-to-peer serving architecture that independently scales prefill, decode, and caching; a large-scale expert parallelism strategy supporting EP320 via efficient UB-based token dispatch; and hardware-aware optimizations including specialized operators, microbatch-based pipelining, and INT8 quantization. Evaluation with the DeepSeek-R1 model shows CloudMatrix-Infer achieves state-of-the-art efficiency: prefill throughput of 6,688 tokens/s per NPU and decode throughput of 1,943 tokens/s per NPU (<50 ms TPOT). It effectively balances throughput and latency, sustaining 538 tokens/s even under stringent 15 ms latency constraints, while INT8 quantization maintains model accuracy across benchmarks.

Large Language Models (LLMs) are increasingly deployed in both latency-sensitive online services and cost-sensitive offline workloads. Co-locating these workloads on shared serving instances can improve resource utilization, but directly applying this approach to Prefill/Decode (P/D) disaggregated systems introduces severe load imbalance, as fluctuating request mixes alter the intrinsic P/D ratio. Existing dynamic adjustment techniques cannot keep up with the bursty traffic patterns of online services. We propose a latency-constraint disaggregated architecture, which separates cluster resources into latency-strict and latency-relaxed pools based on task latency requirements. This design enables flexible placement of offline decode tasks, mitigating P/D imbalance while preserving online performance. To fully exploit this flexibility, we propose (1) a bottleneck-based scheduler guided by a Roofline-based performance model for performance bottleneck based scheduling, and (2) a fast preemption mechanism that strictly enforces Service Level Objectives (SLOs) for online requests. Experiments on real-world traces show that compared to existing offline system approaches, our method improves offline throughput by up to 3x, while maintaining online request SLOs.

Large Language Models (LLMs) have rapidly advanced, with Gemini-3-Pro setting a new performance milestone. In this work, we explore collective intelligence as an alternative to monolithic scaling, and demonstrate that open-source LLMs' collaboration can surpass Gemini-3-Pro. We first revisit LLM routing and aggregation at scale and identify three key bottlenecks: (1) current train-free routers are limited by a query-based paradigm focusing solely on textual similarity; (2) recent aggregation methods remain largely static, failing to select appropriate aggregators for different tasks;(3) the complementarity of routing and aggregation remains underutilized. To address these problems, we introduce JiSi, a novel framework designed to release the full potential of LLMs' collaboration through three innovations: (1) Query-Response Mixed Routing capturing both semantic information and problem difficulty; (2) Support-Set-based Aggregator Selection jointly evaluating the aggregation and domain capacity of aggregators; (3) Adaptive Routing-Aggregation Switch dynamically leveraging the advantages of routing and aggregation. Comprehensive experiments on nine benchmarks demonstrate that JiSi can surpass Gemini-3-Pro with only 47% costs by orchestrating ten open-source LLMs, while outperforming mainstream baselines. It suggests that collective intelligence represents a novel path towards Artificial General Intelligence (AGI).

The rapid adoption of large language models (LLMs) has created significant challenges for efficient inference at scale. Unlike traditional workloads, LLM inference is constrained by both computation and the memory overhead of key-value (KV) caching, which accelerates decoding but quickly exhausts GPU memory. In this paper, we introduce the first queueing-theoretic framework that explicitly incorporates both computation and GPU memory constraints into the analysis of LLM inference. Based on this framework, we derive rigorous stability and instability conditions that determine whether an LLM inference service can sustain incoming demand without unbounded queue growth. This result offers a powerful tool for system deployment, potentially addressing the core challenge of GPU provisioning. By combining an estimated request arrival rate with our derived stable service rate, operators can calculate the necessary cluster size to avoid both costly over-purchasing and performance-violating under-provisioning. We further validate our theoretical predictions through extensive experiments in real GPU production environments. Our results show that the predicted stability conditions are highly accurate, with deviations typically within 10%.

Disaggregated inference has become an essential framework that separates the prefill (P) and decode (D) stages in large language model inference to improve throughput. However, the KV cache transfer faces significant delays between prefill and decode nodes. The block-wise calling method and discontinuous KV cache memory allocation increase the number of calls to the transmission kernel. Additionally, existing frameworks often fix the roles of P and D nodes, leading to computational imbalances. In this paper, we propose FlowKV, a novel disaggregated inference framework, which reduces the average transmission latency of KV cache by 96%, from 0.944s to 0.053s, almost eliminating the transfer time relative to the total request latency by optimizing the KV cache transfer. FlowKV introduces the Load-Aware Scheduler for balanced request scheduling and flexible PD node allocation. This design maximizes hardware resource utilization, achieving peak system throughput across various scenarios, including normal, computational imbalance, and extreme overload conditions. Experimental results demonstrate that FlowKV significantly accelerates inference by 15.2%-48.9% on LongBench dataset compared to the baseline and supports applications with heterogeneous GPUs.

Multi-turn LLM serving accumulates dialogue history whose Key-Value (KV) cache grows with every turn and every user, quickly exceeding the model weights themselves and making memory -- not compute -- the binding constraint on throughput. Non-uniform KV compression, which allocates heterogeneous budgets across attention heads, preserves accuracy far better than uniform schemes, yet remains impractical: modern serving stacks assume identical KV lengths across heads, so heterogeneity traps freed memory as page fragmentation, spends up to 25% of prefill time reclaiming scattered pages, and skews GPU workloads that inflate decode latency by up to 1.7times or burn 15--20% of each decode step on re-planning. We observe that this heterogeneity need not be discovered at runtime: head-wise retention follows a two-level structural regularity -- an input-invariant head ranking with narrowly bounded per-head ratios -- that can be calibrated offline from as few as 50 samples. Building on this insight, we present Tangram, a serving framework that statically resolves what prior systems handle dynamically: Budget Reservation fixes each head's post-compression footprint at scheduling time, eliminating page reclamation; Ragged Paging clusters similar-budget heads into independent page tables, turning fragmentation into reclaimable memory; and Ahead-of-Time Load Balancing precomputes balanced GPU partitions with zero runtime planning. Implemented on vLLM, Tangram serves as a drop-in substrate for existing non-uniform compression methods, matching their accuracy while improving end-to-end throughput by up to 2.6times over the full-KV baseline. Our implementation is publicly available at https://github.com/aiha-lab/TANGRAM.

Recently, various distributed strategies for large language model training have been proposed. However, these methods provided limited solutions for the trade-off between memory consumption and communication cost. In this paper, we rethink the impact of memory consumption and communication costs on the training speed of large language models, and propose a memory-communication balanced strategy set Partial Redundancy Optimizer (PaRO). PaRO provides comprehensive options which reduces the amount and frequency of inter-group communication with minor memory redundancy by fine-grained sharding strategy, thereby improving the training efficiency in various training scenarios. Additionally, we propose a Hierarchical Overlapping Ring (HO-Ring) communication topology to enhance communication efficiency between nodes or across switches in large language model training. Our experiments demonstrate that PaRO significantly improves training throughput by 1.19x-2.50x compared to the SOTA method and achieves a near-linear scalability. The HO-Ring algorithm improves communication efficiency by 36.5% compared to the traditional Ring algorithm.

In this work, we aim to simultaneously enhance the effectiveness and efficiency of Mixture-of-Experts (MoE) methods. To achieve this, we propose MoE++, a general and heterogeneous MoE framework that integrates both Feed-Forward Network~(FFN) and zero-computation experts. Specifically, we introduce three types of zero-computation experts: the zero expert, copy expert, and constant expert, which correspond to discard, skip, and replace operations, respectively. This design offers three key advantages: (i) Low Computing Overhead: Unlike the uniform mixing mechanism for all tokens within vanilla MoE, MoE++ allows each token to engage with a dynamic number of FFNs, be adjusted by constant vectors, or even skip the MoE layer entirely. (ii) High Performance: By enabling simple tokens to utilize fewer FFN experts, MoE++ allows more experts to focus on challenging tokens, thereby unlocking greater performance potential than vanilla MoE. (iii) Deployment Friendly: Given that zero-computation experts have negligible parameters, we can deploy all zero-computation experts on each GPU, eliminating the significant communication overhead and expert load imbalance associated with FFN experts distributed across different GPUs. Moreover, we leverage gating residuals, enabling each token to consider the pathway taken in the previous layer when selecting the appropriate experts. Extensive experimental results demonstrate that MoE++ achieves better performance while delivering 1.1-2.1x expert forward throughput compared to a vanilla MoE model of the same size, which lays a solid foundation for developing advanced and efficient MoE-related models.

Scaling long-context ability is essential for Large Language Models (LLMs). To amortize the memory consumption across multiple devices in long-context training, inter-data partitioning (a.k.a. Data Parallelism) and intra-data partitioning (a.k.a. Context Parallelism) are commonly used. Current training frameworks predominantly treat the two techniques as orthogonal, and establish static communication groups to organize the devices as a static mesh (e.g., a 2D mesh). However, the sequences for LLM training typically vary in lengths, no matter for texts, multi-modalities or reinforcement learning. The mismatch between data heterogeneity and static mesh causes redundant communication and imbalanced computation, degrading the training efficiency. In this work, we introduce ByteScale, an efficient, flexible, and scalable LLM training framework for large-scale mixed training of long and short sequences. The core of ByteScale is a novel parallelism strategy, namely Hybrid Data Parallelism (HDP), which unifies the inter- and intra-data partitioning with a dynamic mesh design. In particular, we build a communication optimizer, which eliminates the redundant communication for short sequences by data-aware sharding and dynamic communication, and further compresses the communication cost for long sequences by selective offloading. Besides, we also develop a balance scheduler to mitigate the imbalanced computation by parallelism-aware data assignment. We evaluate ByteScale with the model sizes ranging from 7B to 141B, context lengths from 256K to 2048K, on a production cluster with more than 12,000 GPUs. Experiment results show that ByteScale outperforms the state-of-the-art training system by up to 7.89x.

We present Nemotron-Labs-3-Puzzle-75B-A9B, a compressed variant of Nemotron-3-Super optimized for interactive deployment. We designed the model to maximize server throughput under high user throughput constraints. In interactive serving workloads on a single 8xB200 node, Puzzle-75B-A9B achieves approximately 2x higher server throughput than Nemotron-3-Super at matched user throughput constraints. In ultra-long-context deployment on a single H100 GPU, the compressed model increases 1M-token concurrency from 1 request to 8 requests. Puzzle-75B-A9B is constructed using a multi-stage pipeline that combines the Iterative Puzzle compression framework with knowledge distillation, reinforcement learning, quantization, and a Multi-Token Prediction head. The compression process jointly optimizes heterogeneous MoE pruning, active parameter budget, and Mamba pruning to improve inference efficiency while preserving model quality. We evaluate Puzzle-75B-A9B on a broad suite of reasoning, coding, multilingual, long-context, and agentic benchmarks. Despite substantial compression, the model retains strong downstream accuracy relative to the parent model across a wide range of tasks. These results demonstrate that large hybrid MoE models can be substantially optimized for deployment efficiency while maintaining strong downstream capability.

Mixture-of-Experts (MoE) model architecture has emerged as a promising solution for scaling transformer models efficiently, offering sparse activation that reduces computational costs while increasing model capacity. However, as MoE models scale, they need to be distributed across GPU devices, thus face critical performance bottlenecks due to their large memory footprint. Expert parallelism distributes experts across GPUs, however, faces key challenges including an unbalanced token routing and expert activation, resulting in communication tail latency and processing inefficiencies. While existing solutions address some of these issues, they fail to resolve the dual challenges of load imbalance and communication skew. The imbalance in token processing load across experts causes uneven processing times on different GPUs, while communication skew between GPUs leads to unbalanced inter-GPU data transfers. These factors degrade the performance of MoE models by increasing tail latency and reducing overall throughput. To address these limitations, we propose an Integer Linear Programming (ILP) formulation to optimize expert placement by jointly considering token load, communication, and computation costs. We exploit the property that there is a token routing dependency across layers, where tokens routed to a specific expert in one layer are likely to be routed to a limited set of experts in the subsequent layer. Our solution, MoETuner, offers an optimal expert-to-GPU assignment that minimizes inter-GPU token routing costs and balances token processing across devices, thereby reducing tail latency and end-to-end execution time. Experimental results demonstrate 9.3% and 17.5% of end-to-end speedups for single-node and multi-node inference respectively, showcasing the potential of our ILP-based optimization for offering expert parallel solutions for next-generation MoEs.

Mixture-of-Experts (MoE) enables efficient scaling of large language models (LLMs) with sparsely activated experts during inference. To effectively deploy large MoE models on memory-constrained devices, many systems introduce *expert offloading* that caches a subset of experts in fast memory, leaving others on slow memory to run on CPU or load on demand. While some research has exploited the locality of expert activations, where consecutive tokens activate similar experts, the degree of this **local routing consistency** varies across models and remains understudied. In this paper, we propose two metrics to measure local routing consistency of MoE models: (1) **Segment Routing Best Performance (SRP)**, which evaluates how well a fixed group of experts can cover the needs of a segment of tokens, and (2) **Segment Cache Best Hit Rate (SCH)**, which measures the optimal segment-level cache hit rate under a given cache size limit. We analyzed 20 MoE LLMs with diverse sizes and architectures and found that models that apply MoE on every layer and do not use shared experts exhibit the highest local routing consistency. We further showed that domain-specialized experts contribute more to routing consistency than vocabulary-specialized ones, and that most models can balance between cache effectiveness and efficiency with cache sizes approximately 2x the active experts. These findings pave the way for memory-efficient MoE design and deployment without compromising inference speed. We publish the code for replicating experiments at https://github.com/ljcleo/moe-lrc .

Increasing demand for Large Language Models (LLMs) services imposes substantial deployment and computation costs on providers. LLM routing offers a cost-efficient solution by directing queries to the optimal LLM based on model and query features. However, existing works primarily focus on offline scenarios and struggle to adapt to online settings with high query volume and constrained token budgets. In this work, we introduce the first training-free algorithm for online routing scenarios. Our algorithm leverages approximate nearest neighbor search to efficiently estimate query features and performs a one-time optimization over a small set of initial queries to learn a routing strategy that guides future routing. We provide theoretical guarantees demonstrating that our algorithm achieves a competitive ratio of 1 - o(1) under natural assumptions, which is further validated by extensive experiments across 3 benchmark datasets and 8 baselines, showing an average improvement of 3.55times in overall performance, 1.85times in cost efficiency, and nearly 4.25times in throughput. Our code is available at https://github.com/fzwark/PORT.

While scaling laws for recommendation models have gained significant traction, existing architectures such as Wukong, HiFormer and DHEN, often struggle with sub-optimal designs and hardware under-utilization, limiting their practical scalability. Our previous TokenMixer architecture (introduced in RankMixer paper) addressed effectiveness and efficiency by replacing self-attention with a ightweight token-mixing operator; however, it faced critical bottlenecks in deeper configurations, including sub-optimal residual paths, vanishing gradients, incomplete MoE sparsification and constrained scalability. In this paper, we propose TokenMixer-Large, a systematically evolved architecture designed for extreme-scale recommendation. By introducing a mixing-and-reverting operation, inter-layer residuals and the auxiliary loss, we ensure stable gradient propagation even as model depth increases. Furthermore, we incorporate a Sparse Per-token MoE to enable efficient parameter expansion. TokenMixer-Large successfully scales its parameters to 7-billion and 15-billion on online traffic and offline experiments, respectively. Currently deployed in multiple scenarios at ByteDance, TokenMixer-Large has achieved significant offline and online performance gains, delivering an increase of +1.66\% in orders and +2.98\% in per-capita preview payment GMV for e-commerce, improving ADSS by +2.0\% in advertising and achieving a +1.4\% revenue growth for live streaming.

In prefill-decode (PD) disaggregated LLM serving, each request is assigned to a decode worker after prefill. Existing decode routers balance only load; for mixture-of-experts (MoE) models this is incomplete: equally loaded workers can differ in latency, since each decode step loads the weights of every distinct expert its batch activates. We present ELDR, an expert-locality-aware decode router for PD-disaggregated MoE serving. From a request's prefill expert activations, ELDR builds an expert signature predicting the experts it will activate during generation. Offline, balanced K-means partitions signature space across decode workers; online, locality-band routing sends each request to the least-loaded worker among those best matching its signature. A signature cache, co-indexed with the KV cache at KV-block granularity, keeps signatures exact under prefix caching. Implemented in vLLM and evaluated on deployments of up to 40 GPUs, ELDR reduces median TPOT by 5.9-13.9% over the strongest of four load-balancing baselines across three MoE models and two workloads, with model outputs unchanged.

Sparse Mixture of Experts (MoE) models, while outperforming dense Large Language Models (LLMs) in terms of performance, face significant deployment challenges during inference due to their high memory demands. Existing offloading techniques, which involve swapping activated and idle experts between the GPU and CPU, often suffer from rigid expert caching mechanisms. These mechanisms fail to adapt to dynamic routing, leading to inefficient cache utilization, or incur prohibitive costs for prediction training. To tackle these inference-specific challenges, we introduce ExpertFlow, a comprehensive system specifically designed to enhance inference efficiency by accommodating flexible routing and enabling efficient expert scheduling between CPU and GPU. This reduces overhead and boosts system performance. Central to our approach is a predictive routing path-based offloading mechanism that utilizes a lightweight predictor to accurately forecast routing paths before computation begins. This proactive strategy allows for real-time error correction in expert caching, significantly increasing cache hit ratios and reducing the frequency of expert transfers, thereby minimizing I/O overhead. Additionally, we implement a dynamic token scheduling strategy that optimizes MoE inference by rearranging input tokens across different batches. This method not only reduces the number of activated experts per batch but also improves computational efficiency. Our extensive experiments demonstrate that ExpertFlow achieves up to 93.72\% GPU memory savings and enhances inference speed by 2 to 10 times compared to baseline methods, highlighting its effectiveness and utility as a robust solution for resource-constrained inference scenarios.

Many LLM tasks are performed in large batches or even offline, and the performance indictor for which is throughput. These tasks usually show the characteristic of prefix sharing, where different prompt input can partially show the common prefix. However, the existing LLM inference engines tend to optimize the streaming requests and show limitations of supporting the large batched tasks with the prefix sharing characteristic. The existing solutions use the LRU-based cache to reuse the KV context of common prefix. The KV context that is about to be reused may prematurely be evicted with the implicit cache management. Even if not evicted, the lifetime of the shared KV context is extended since requests sharing the same context are not scheduled together, resulting in larger memory usage. These streaming oriented systems schedule the requests in the first-come-first-serve or similar order. As a result, the requests with larger ratio of decoding steps may be scheduled too late to be able to mix with the prefill chunks to increase the hardware utilization. Besides, the token and request number based batching can limit the size of token-batch, which keeps the GPU from saturating for the iterations dominated by decoding tokens. We propose BatchLLM to address the above problems. BatchLLM explicitly identifies the common prefixes globally. The requests sharing the same prefix will be scheduled together to reuse the KV context the best, which also shrinks the lifetime of common KV memory. BatchLLM reorders the requests and schedules the requests with larger ratio of decoding first to better mix the decoding tokens with the latter prefill chunks and applies memory-centric token batching to enlarge the token-batch sizes, which helps to increase the GPU utilization. Extensive evaluation shows that BatchLLM outperforms vLLM by 1.1x to 2x on a set of microbenchmarks and two typical industry workloads.

Scaling laws enable the optimal selection of data amount and language model size, yet the impact of the data unit, the token, on this relationship remains underexplored. In this work, we systematically investigate how the information granularity of tokens, controlled by the compression rate (i.e., average bytes of text per token), affects scaling trends. We train 988 latent tokenized models (BLT) ranging from 50M to 7B parameters that enable setting the desired compression rate. This flexibility allows us to study the role of compression rate well beyond 4.57 bytes per token obtained with a popular BPE tokenizer. Our experiments reveal that in compute-optimal configurations, model parameter counts scale proportionally to data size measured in bytes, not in tokens as commonly perceived (Kaplan et al., 2020; Hoffmann et al., 2022). Furthermore, we discover that the optimal compression rate differs from the one obtained with BPE and decreases with compute. These findings generalize to both latent and subword tokenization, as well as to languages other than English, guiding language model developers on tokenization scheme selection for maximal compute efficiency.

Video Diffusion Transformers process long spatio-temporal sequences, making self-attention the main bottleneck in high-resolution video generation. Training-free sparse attention reduces this cost, but adaptive Top-p routing creates uneven per-head workloads under multi-GPU sequence parallelism. The resulting workload heterogeneity turns sparse attention into a rank-level straggler problem. We present , a training-free sparse-attention system that improves the distributed execution efficiency of adaptive sparse attention under multi-GPU sequence parallelism. uses Top-p routing, a Top-k safety floor, and video-aware block organization as the sparse-routing frontend, then repairs the materialized mask at runtime. Runtime Load Balancing migrates a small number of heavy heads via P2P communication to shorten the current critical path. Slack-Aware Sparse Augmentation fills residual non-critical-rank slack with additional high-value blocks, while overlap hides scheduling and migration overhead behind existing computation. On step-distilled Wan2.2 I2V, reduces average load imbalance from 1.34 to 1.08 and delivers a 4.41times attention speedup over FlashAttention, while achieving a 2.02--2.11times DiT inference speedup with competitive video quality.

The mixture of Expert (MoE) parallelism is a recent advancement that scales up the model size with constant computational cost. MoE selects different sets of parameters (i.e., experts) for each incoming token, resulting in a sparsely-activated model. Despite several successful applications of MoE, its training efficiency degrades significantly as the number of experts increases. The routing stage in MoE relies on the efficiency of the All2All communication collective, which suffers from network congestion and has poor scalability. To mitigate these issues, we introduce SMILE, which exploits heterogeneous network bandwidth and splits a single-step routing into bi-level routing. Our experimental results show that the proposed method obtains a 2.5x speedup over Switch Transformer in terms of pretraining throughput on the Colossal Clean Crawled Corpus without losing any convergence speed.

This paper releases and analyzes a month-long trace of 85 billion user requests and 11.9 million cold starts from Huawei's serverless cloud platform. Our analysis spans workloads from five data centers. We focus on cold starts and provide a comprehensive examination of the underlying factors influencing the number and duration of cold starts. These factors include trigger types, request synchronicity, runtime languages, and function resource allocations. We investigate components of cold starts, including pod allocation time, code and dependency deployment time, and scheduling delays, and examine their relationships with runtime languages, trigger types, and resource allocation. We introduce pod utility ratio to measure the pod's useful lifetime relative to its cold start time, giving a more complete picture of cold starts, and see that some pods with long cold start times have longer useful lifetimes. Our findings reveal the complexity and multifaceted origins of the number, duration, and characteristics of cold starts, driven by differences in trigger types, runtime languages, and function resource allocations. For example, cold starts in Region 1 take up to 7 seconds, dominated by dependency deployment time and scheduling. In Region 2, cold starts take up to 3 seconds and are dominated by pod allocation time. Based on this, we identify opportunities to reduce the number and duration of cold starts using strategies for multi-region scheduling. Finally, we suggest directions for future research to address these challenges and enhance the performance of serverless cloud platforms. Our datasets and code are available here https://github.com/sir-lab/data-release

The deployment and scaling of large language models (LLMs) have become critical as they permeate various applications, demanding high-throughput and low-latency serving systems. Existing frameworks struggle to balance these requirements, especially for workloads with long prompts. This paper introduces DeepSpeed-FastGen, a system that employs Dynamic SplitFuse, a novel prompt and generation composition strategy, to deliver up to 2.3x higher effective throughput, 2x lower latency on average, and up to 3.7x lower (token-level) tail latency, compared to state-of-the-art systems like vLLM. We leverage a synergistic combination of DeepSpeed-MII and DeepSpeed-Inference to provide an efficient and easy-to-use serving system for LLMs. DeepSpeed-FastGen's advanced implementation supports a range of models and offers both non-persistent and persistent deployment options, catering to diverse user scenarios from interactive sessions to long-running applications. We present a detailed benchmarking methodology, analyze the performance through latency-throughput curves, and investigate scalability via load balancing. Our evaluations demonstrate substantial improvements in throughput and latency across various models and hardware configurations. We discuss our roadmap for future enhancements, including broader model support and new hardware backends. The DeepSpeed-FastGen code is readily available for community engagement and contribution.

Standard transport protocols like TCP operate as a blind, FIFO conveyor belt for data, a model that is increasingly suboptimal for latency-sensitive and interactive applications. This paper challenges this model by introducing CATS (Conductor-driven Asymmetric Transport Scheme), a framework that provides TCP with the semantic awareness necessary to prioritize critical content. By centralizing scheduling intelligence in a transport-native "Conductor", CATS significantly improves user-perceived performance by delivering essential data first. This architecture directly confronts a cascade of historical performance workarounds and their limitations, including the high overhead of parallel connections in HTTP/1.1, the transport-layer Head-of-Line blocking in HTTP/2, and the observed implementation heterogeneity of prioritization in HTTP/3 over QUIC. Built upon TCP BBR, our ns-3 implementation demonstrates this principle by reducing the First Contentful Paint by over 78% in a representative webpage download configured as a deliberate worst-case scenario, with no penalty to total page load time compared to the baseline.

Training transformer models requires substantial GPU compute and memory resources. In homogeneous clusters, distributed strategies allocate resources evenly, but this approach is inefficient for heterogeneous clusters, where GPUs differ in power and memory. As high-end GPUs are costly and limited in availability, heterogeneous clusters with diverse GPU types are becoming more common. Existing methods attempt to balance compute across GPUs based on capacity but often underutilize compute due to memory constraints. We present Cephalo, a system that optimizes compute and memory usage by decoupling compute distribution from training state assignment. Cephalo outperforms state-of-the-art methods by achieving significantly higher training throughput while supporting larger models and batch sizes.

Mixture-of-Experts (MoE) has emerged as a promising paradigm for foundation models due to its efficient and powerful scalability. In this work, we present Sigma-MoE-Tiny, an MoE language model that achieves the highest sparsity compared to existing open-source models. Sigma-MoE-Tiny employs fine-grained expert segmentation with up to 96 experts per layer, while activating only one expert for each token, resulting in 20B total parameters with just 0.5B activated. The major challenge introduced by such extreme sparsity lies in expert load balancing. We find that the widely-used load balancing loss tends to become ineffective in the lower layers under this setting. To address this issue, we propose a progressive sparsification schedule aiming to balance expert utilization and training stability. Sigma-MoE-Tiny is pre-trained on a diverse and high-quality corpus, followed by post-training to further unlock its capabilities. The entire training process remains remarkably stable, with no occurrence of irrecoverable loss spikes. Comprehensive evaluations reveal that, despite activating only 0.5B parameters, Sigma-MoE-Tiny achieves top-tier performance among counterparts of comparable or significantly larger scale. In addition, we provide an in-depth discussion of load balancing in highly sparse MoE models, offering insights for advancing sparsity in future MoE architectures. Project page: https://qghuxmu.github.io/Sigma-MoE-Tiny Code: https://github.com/microsoft/ltp-megatron-lm

Mixture-of-Experts layers achieve compute efficiency through weight sparsity: each token activates only a subset of experts. Data sparsity, where each expert processes only a subset of tokens, offers a complementary axis. Expert-choice routing implements data sparsity directly but violates causality in autoregressive models, creating train-inference mismatch. We recover data sparsity within causal token-choice MoE by leveraging zero-compute (null) experts within the routing pool. When a token routes to null experts, those slots consume no compute. The standard load balancing objective trains the model to uniformly use all experts (real and null) therefore creating data sparsity in expectation without the causality violations. We evaluate on vision-language model training, where data heterogeneity is pronounced: vision encoders produce many low-information tokens while text tokens are denser. At matched expected FLOPs, composing weight and data sparsity yields a more compute-efficient frontier than weight sparsity alone, with gains in training loss and downstream performance. The model learns implicit modality-aware allocation, routing vision tokens to null experts more aggressively than text, without explicit modality routing.

Sparsely-gated mixture-of-experts (MoE) has been widely adopted to scale deep learning models to trillion-plus parameters with fixed computational cost. The algorithmic performance of MoE relies on its token routing mechanism that forwards each input token to the right sub-models or experts. While token routing dynamically determines the amount of expert workload at runtime, existing systems suffer inefficient computation due to their static execution, namely static parallelism and pipelining, which does not adapt to the dynamic workload. We present Flex, a highly scalable stack design and implementation for MoE with dynamically adaptive parallelism and pipelining. Flex designs an identical layout for distributing MoE model parameters and input data, which can be leveraged by all possible parallelism or pipelining methods without any mathematical inequivalence or tensor migration overhead. This enables adaptive parallelism/pipelining optimization at zero cost during runtime. Based on this key design, Flex also implements various MoE acceleration techniques. Aggregating all techniques, Flex finally delivers huge speedup at any scale -- 4.96x and 5.75x speedup of a single MoE layer over 16 and 2,048 A100 GPUs, respectively, over the previous state-of-the-art. Our evaluation shows that Flex efficiently and effectively runs a real-world MoE-based model named SwinV2-MoE, built upon Swin Transformer V2, a state-of-the-art computer vision architecture. On efficiency, Flex accelerates SwinV2-MoE, achieving up to 1.55x and 2.11x speedup in training and inference over Fairseq, respectively. On effectiveness, the SwinV2-MoE model achieves superior accuracy in both pre-training and down-stream computer vision tasks such as COCO object detection than the counterpart dense model, indicating the readiness of Flex for end-to-end real-world model training and inference.

The rapid advancement of Large Language Models (LLMs) has driven the need for more efficient serving strategies. In this context, efficiency refers to the proportion of requests that meet their Service Level Objectives (SLOs), particularly for Time To First Token (TTFT) and Time Between Tokens (TBT). However, existing systems often prioritize one metric at the cost of the other. We present Ascendra, an LLM serving system designed to meet both TTFT and TBT SLOs simultaneously. The core insight behind Ascendra is that a request's urgency evolves as it approaches its deadline. To leverage this, Ascendra partitions GPU resources into two types of instances: low-priority and high-priority. Low-priority instances maximize throughput by processing requests out of arrival order, but at the risk of request starvation. To address this, Ascendra employs a performance model to predict requests at risk of missing their SLOs and proactively offloads them to high-priority instances. High-priority instances are optimized for low-latency execution and handle urgent requests nearing their deadlines. This partitioned architecture enables Ascendra to effectively balance high throughput and low latency. Extensive evaluation shows that Ascendra improves system throughput by up to 1.7x compared to vLLM and Sarathi-Serve while meeting both TTFT and TBT SLOs.

Large Language Models (LLMs) for Generative AI have achieved remarkable progress, evolving into sophisticated and versatile tools widely adopted across various domains and applications. However, the substantial memory overhead caused by their vast number of parameters, combined with the high computational demands of the attention mechanism, poses significant challenges in achieving low latency and high throughput for LLM inference services. Recent advancements, driven by groundbreaking research, have significantly accelerated progress in this field. This paper provides a comprehensive survey of these methods, covering fundamental instance-level approaches, in-depth cluster-level strategies, emerging scenario directions, and other miscellaneous but important areas. At the instance level, we review model placement, request scheduling, decoding length prediction, storage management, and the disaggregation paradigm. At the cluster level, we explore GPU cluster deployment, multi-instance load balancing, and cloud service solutions. For emerging scenarios, we organize the discussion around specific tasks, modules, and auxiliary methods. To ensure a holistic overview, we also highlight several niche yet critical areas. Finally, we outline potential research directions to further advance the field of LLM inference serving.

Large language model (LLM) routing assigns each query to the most suitable model from an ensemble. We introduce LLMRouterBench, a large-scale benchmark and unified framework for LLM routing. It comprises over 400K instances from 21 datasets and 33 models. Moreover, it provides comprehensive metrics for both performance-oriented routing and performance-cost trade-off routing, and integrates 10 representative routing baselines. Using LLMRouterBench, we systematically re-evaluate the field. While confirming strong model complementarity-the central premise of LLM routing-we find that many routing methods exhibit similar performance under unified evaluation, and several recent approaches, including commercial routers, fail to reliably outperform a simple baseline. Meanwhile, a substantial gap remains to the Oracle, driven primarily by persistent model-recall failures. We further show that backbone embedding models have limited impact, that larger ensembles exhibit diminishing returns compared to careful model curation, and that the benchmark also enables latency-aware analysis. All code and data are available at https://github.com/ynulihao/LLMRouterBench.

Attention efficiency is critical to large language model (LLM) inference. While prior advances optimize attention execution for individual requests (e.g., FlashAttention), production LLM serving relies on batching requests with highly heterogeneous sequence lengths for high serving throughput. This mismatch induces severe computation and I/O imbalance, exacerbates stragglers, and underutilizes GPU resources. We present PackInfer, a kernel-level attention framework that enables compute- and I/O-aware execution for heterogeneous batched inference. PackInfer orchestrates batched requests into load-balanced execution groups, effectively saturating GPU utilization by packing multiple requests into unified kernel launches. By constructing attention kernels directly over packed query-key regions, PackInfer eliminates redundant computation and balances thread-block execution. It then incorporates I/O-aware grouping that co-locates shared-prefix requests and reorganizes KV caches into group-contiguous layouts, reducing memory fragmentation and redundant data movement as generation evolves. Evaluations on real-world workloads show that PackInfer reduces inference latency by 13.0-20.1%, and improves throughput by 20% compared to the state-of-the-art FlashAttention.

Text-to-SQL automatically translates natural language queries to SQL, allowing non-technical users to retrieve data from databases without specialized SQL knowledge. Despite the success of advanced LLM-based Text-to-SQL approaches on leaderboards, their unsustainable computational costs--often overlooked--stand as the "elephant in the room" in current leaderboard-driven research, limiting their economic practicability for real-world deployment and widespread adoption. To tackle this, we exploratively propose EllieSQL, a complexity-aware routing framework that assigns queries to suitable SQL generation pipelines based on estimated complexity. We investigate multiple routers to direct simple queries to efficient approaches while reserving computationally intensive methods for complex cases. Drawing from economics, we introduce the Token Elasticity of Performance (TEP) metric, capturing cost-efficiency by quantifying the responsiveness of performance gains relative to token investment in SQL generation. Experiments show that compared to always using the most advanced methods in our study, EllieSQL with the Qwen2.5-0.5B-DPO router reduces token use by over 40% without compromising performance on Bird development set, achieving more than a 2x boost in TEP over non-routing approaches. This not only advances the pursuit of cost-efficient Text-to-SQL but also invites the community to weigh resource efficiency alongside performance, contributing to progress in sustainable Text-to-SQL.

Code large language models (Code LLMs) are powerful but costly to train, with scaling laws predicting performance from model size, data, and compute. However, different programming languages (PLs) have varying impacts during pre-training that significantly affect base model performance, leading to inaccurate performance prediction. Besides, existing works focus on language-agnostic settings, neglecting the inherently multilingual nature of modern software development. Therefore, it is first necessary to investigate the scaling laws of different PLs, and then consider their mutual influences to arrive at the final multilingual scaling law. In this paper, we present the first systematic exploration of scaling laws for multilingual code pre-training, conducting over 1000+ experiments (Equivalent to 336,000+ H800 hours) across multiple PLs, model sizes (0.2B to 14B parameters), and dataset sizes (1T tokens). We establish comprehensive scaling laws for code LLMs across multiple PLs, revealing that interpreted languages (e.g., Python) benefit more from increased model size and data than compiled languages (e.g., Rust). The study demonstrates that multilingual pre-training provides synergistic benefits, particularly between syntactically similar PLs. Further, the pre-training strategy of the parallel pairing (concatenating code snippets with their translations) significantly enhances cross-lingual abilities with favorable scaling properties. Finally, a proportion-dependent multilingual scaling law is proposed to optimally allocate training tokens by prioritizing high-utility PLs (e.g., Python), balancing high-synergy pairs (e.g., JavaScript-TypeScript), and reducing allocation to fast-saturating languages (Rust), achieving superior average performance across all PLs compared to uniform distribution under the same compute budget.

This paper introduces bucket calculus, a novel mathematical framework that fundamentally transforms the computational complexity landscape of parallel machine scheduling optimization. We address the strongly NP-hard problem P2|r_j|C_{max} through an innovative adaptive temporal discretization methodology that achieves exponential complexity reduction from O(T^n) to O(B^n) where B ll T, while maintaining near-optimal solution quality. Our bucket-indexed mixed-integer linear programming (MILP) formulation exploits dimensional complexity heterogeneity through precision-aware discretization, reducing decision variables by 94.4\% and achieving a theoretical speedup factor 2.75 times 10^{37} for 20-job instances. Theoretical contributions include partial discretization theory, fractional bucket calculus operators, and quantum-inspired mechanisms for temporal constraint modeling. Empirical validation on instances with 20--400 jobs demonstrates 97.6\% resource utilization, near-perfect load balancing (σ/μ= 0.006), and sustained performance across problem scales with optimality gaps below 5.1\%. This work represents a paradigm shift from fine-grained temporal discretization to multi-resolution precision allocation, bridging the fundamental gap between exact optimization and computational tractability for industrial-scale NP-hard scheduling problems.

Ad-load balancing is a critical challenge in online advertising systems, particularly in the context of social media platforms, where the goal is to maximize user engagement and revenue while maintaining a satisfactory user experience. This requires the optimization of conflicting objectives, such as user satisfaction and ads revenue. Traditional approaches to ad-load balancing rely on static allocation policies, which fail to adapt to changing user preferences and contextual factors. In this paper, we present an approach that leverages off-policy learning and evaluation from logged bandit feedback. We start by presenting a motivating analysis of the ad-load balancing problem, highlighting the conflicting objectives between user satisfaction and ads revenue. We emphasize the nuances that arise due to user heterogeneity and the dependence on the user's position within a session. Based on this analysis, we define the problem as determining the optimal ad-load for a particular feed fetch. To tackle this problem, we propose an off-policy learning framework that leverages unbiased estimators such as Inverse Propensity Scoring (IPS) and Doubly Robust (DR) to learn and estimate the policy values using offline collected stochastic data. We present insights from online A/B experiments deployed at scale across over 80 million users generating over 200 million sessions, where we find statistically significant improvements in both user satisfaction metrics and ads revenue for the platform.

This paper presents a strategy to allocate services on a Cloud system without overloading nodes and maintaining the system stability with minimum cost. We specify an abstract model of cloud resources utilization, including multiple types of resources as well as considerations for the service migration costs. A prototype meta-heuristic load balancer is demonstrated and experimental results are presented and discussed. We also propose a novel genetic algorithm, where population is seeded with the outputs of other meta-heuristic algorithms.

Standard continuous-time generative models rely on monolithic architectures that must navigate vastly different signal regimes, from isotropic noise to intricate data distributions. While scaling model capacity improves performance, deploying a massive network uniformly across the entire generative timeline is inherently inefficient. In this work, we propose Complexity-Balanced Splitting (CBS), a principled framework for temporal capacity allocation that distributes the generative workload across multiple specialized sub-networks. Grounded in function approximation theory and de Boor's equidistribution principle, CBS partitions the diffusion timeline into segments of equal approximation burden, allocating more representational capacity to regions where the generative dynamics are more difficult to model. To estimate this local complexity, we introduce two complementary and tractable monitor functions: a spatial measure based on the flow's Dirichlet energy, and a geometric measure based on the acceleration of the sampling trajectories. Using a lightweight auxiliary model to estimate these complexity profiles, our approach eliminates the need for heuristic temporal splits or computationally expensive search procedures. Extensive evaluation across multiple architectures (SiT, JiT, and UNet) and datasets demonstrates that CBS consistently improves synthesis quality without increasing per-step inference cost. In particular, CBS improves FID by ~35% on SiT-XL with CFG relative to naive temporal partitioning. Project page is available at https://noamissachar.github.io/CBS/.

As the range of applications for Large Language Models (LLMs) continues to grow, the demand for effective serving solutions becomes increasingly critical. Despite the versatility of LLMs, no single model can optimally address all tasks and applications, particularly when balancing performance with cost. This limitation has led to the development of LLM routing systems, which combine the strengths of various models to overcome the constraints of individual LLMs. Yet, the absence of a standardized benchmark for evaluating the performance of LLM routers hinders progress in this area. To bridge this gap, we present RouterBench, a novel evaluation framework designed to systematically assess the efficacy of LLM routing systems, along with a comprehensive dataset comprising over 405k inference outcomes from representative LLMs to support the development of routing strategies. We further propose a theoretical framework for LLM routing, and deliver a comparative analysis of various routing approaches through RouterBench, highlighting their potentials and limitations within our evaluation framework. This work not only formalizes and advances the development of LLM routing systems but also sets a standard for their assessment, paving the way for more accessible and economically viable LLM deployments. The code and data are available at https://github.com/withmartian/routerbench.

Speculative decoding accelerates Large Language Model (LLM) inference by decoupling draft generation from target verification. While recent parallel drafters efficiently propose long token sequences in a single forward pass, they suffer from rapid acceptance decay due to a lack of inter-token dependencies. Furthermore, indiscriminately verifying these extended blocks wastes critical batch capacity on tokens with high rejection risks, severely degrading throughput in high-concurrency serving systems. We introduce DSpark, a speculative decoding framework that unifies high-throughput parallel generation with adaptive, load-aware verification. To maintain draft quality, DSpark utilizes a semi-autoregressive architecture, coupling a parallel backbone with a lightweight sequential module, to introduce intra-block dependency modeling and mitigate suffix decay. To optimize system efficiency, DSpark employs confidence-scheduled verification, dynamically tailoring the verification length for each request based on estimated prefix survival probabilities and engine-specific throughput profiles. On offline benchmarks across diverse domains, DSpark substantially improves the accepted length over state-of-the-art autoregressive and parallel drafters. When deployed within the DeepSeek-V4 serving system under live user traffic, DSpark successfully mitigates verification waste. Compared to the established production baseline (MTP-1), DSpark accelerates per-user generation speeds by 60 to 85 percent at matched throughput levels. More importantly, by preventing severe throughput degradation under strict interactivity constraints, it enables performance tiers that were previously unattainable, shifting the Pareto frontier of our serving system.

The rise of LLMs has driven demand for private serverless deployments, characterized by moderate-scale models and infrequent requests. While existing solutions follow exclusive GPU deployment, we take a step back to explore modern platforms and find that: Emerging CPU architectures with built-in accelerators are capable of serving LLMs but remain underutilized, and both CPUs and GPUs can accommodate multiple LLMs simultaneously. We propose LLM-Mesh, a serverless inference scheme for small-to-mid-sized LLMs that enables elastic sharing across heterogeneous hardware. LLM-Mesh tackles three fundamental challenges: (1) precise, fine-grained compute resource allocation at token-level to handle fluctuating computational demands; (2) a coordinated and forward-looking memory scaling mechanism to detect out-of-memory hazards and reduce operational overhead; and (3) a dual approach that reduces resource fragmentation through proactive preemption and reactive bin-packing. Experimental results on 4 32-core CPUs and 4 A100 GPUs show that LLM-Meshimproves service capacity by 44% - 63% through sharing, while further leveraging CPUs boosts this to 91% - 159%.

With the widespread adoption of Large Language Models (LLMs), serving LLM inference requests has become an increasingly important task, attracting active research advancements. Practical workloads play an essential role in this process: they are critical for motivating and benchmarking serving techniques and systems. However, the existing understanding of real-world LLM serving workloads is limited due to the lack of a comprehensive workload characterization. Prior analyses remain insufficient in scale and scope, thus failing to fully capture intricate workload characteristics. In this paper, we fill the gap with an in-depth characterization of LLM serving workloads collected from our worldwide cloud inference serving service, covering not only language models but also emerging multimodal and reasoning models, and unveiling important new findings in each case. Moreover, based on our findings, we propose ServeGen, a principled framework for generating realistic LLM serving workloads by composing them on a per-client basis. A practical use case in production validates that ServeGen avoids 50% under-provisioning compared to naive workload generation, demonstrating ServeGen's advantage in performance benchmarking. We will open-source ServeGen to foster future research.

Serverless computing has emerged as a compelling new paradigm of cloud computing models in recent years. It promises the user services at large scale and low cost while eliminating the need for infrastructure management. On cloud provider side, flexible resource management is required to meet fluctuating demand. It can be enabled through automated provisioning and deprovisioning of resources. A common approach among both commercial and open source serverless computing platforms is workload-based auto-scaling, where a designated algorithm scales instances according to the number of incoming requests. In the recently evolving serverless framework Knative a request-based policy is proposed, where the algorithm scales resources by a configured maximum number of requests that can be processed in parallel per instance, the so-called concurrency. As we show in a baseline experiment, this predefined concurrency level can strongly influence the performance of a serverless application. However, identifying the concurrency configuration that yields the highest possible quality of service is a challenging task due to various factors, e.g. varying workload and complex infrastructure characteristics, influencing throughput and latency. While there has been considerable research into intelligent techniques for optimizing auto-scaling for virtual machine provisioning, this topic has not yet been discussed in the area of serverless computing. For this reason, we investigate the applicability of a reinforcement learning approach, which has been proven on dynamic virtual machine provisioning, to request-based auto-scaling in a serverless framework. Our results show that within a limited number of iterations our proposed model learns an effective scaling policy per workload, improving the performance compared to the default auto-scaling configuration.

LLM serving systems process heterogeneous query workloads where different categories exhibit different characteristics. Code queries cluster densely in embedding space while conversational queries distribute sparsely. Content staleness varies from minutes (stock data) to months (code patterns). Query repetition patterns range from power-law (code) to uniform (conversation), producing long tail cache hit rate distributions: high-repetition categories achieve 40-60% hit rates while low-repetition or volatile categories achieve 5-15% hit rates. Vector databases must exclude the long tail because remote search costs (30ms) require 15--20% hit rates to break even, leaving 20-30% of production traffic uncached. Uniform cache policies compound this problem: fixed thresholds cause false positives in dense spaces and miss valid paraphrases in sparse spaces; fixed TTLs waste memory or serve stale data. This paper presents category-aware semantic caching where similarity thresholds, TTLs, and quotas vary by query category. We present a hybrid architecture separating in-memory HNSW search from external document storage, reducing miss cost from 30ms to 2ms. This reduction makes low-hit-rate categories economically viable (break-even at 3-5% versus 15-20%), enabling cache coverage across the entire workload distribution. Adaptive load-based policies extend this framework to respond to downstream model load, dynamically adjusting thresholds and TTLs to reduce traffic to overloaded models by 9-17% in theoretical projections.

Large language models (LLMs) propel the prosperity of interactive AI applications showcased by ChatGPT that demand timely response of inference services. However, LLM inference is computation intensive and memory intensive, and improper parameter configuration at LLM platforms may exacerbate the inference time. In this paper, we analyze the impact of LLM output token distribution on the inference queueing delay, where the max-token clipping and the batched inference are considered. By formulating an M/G/1 model, we observe that enforcing a maximum output token limit on a very small fraction of inference requests can significantly reduce the queueing delay, and our model facilitates the selection of the optimal limit. For the batch inference, we model the service process as a bulk queue in which the batch processing time is affected by the batch size and the maximum token size inside this batch jointly. The queueing delays of the batching of all buffered requests (dynamic batching), the batching of constant number of requests (fixed batching), and the batching without intra-batch waiting (elastic batching) are derived. Experimental results show that our mathematical models coincide with the event-driven simulations well.

Parameter-Efficient Fine-Tuning (PEFT) is widely applied as the backend of fine-tuning APIs for large language model (LLM) customization in datacenters. Service providers deploy separate instances for individual PEFT tasks, giving rise to prominent resource inefficiencies, including (1) GPU underutilization from small-scale, PEFT-native operators and (2) device stalls from communication delays and data dependencies in parallelized execution. To address these issues, this paper presents MuxTune, a fine-tuning system that enables resource-efficient concurrent execution of multiple PEFT tasks. The key idea is to multiplex the backbone across independent tasks in a spatial-temporal manner for improved utilization and reduced stalls. Building on flexible, modularized backbone sharing via unified PEFT representations, MuxTune proposes hierarchical co-scheduling scheme with task, operator, and data-level optimizations. Specifically, it fuses tasks through a hybrid of spatial and temporal multiplexing, and orchestrates multi-task operator execution in two-tiered hybrid parallelism. Additionally, MuxTune employs chunk-based data alignment to mitigate inter-task ineffective tokens. Experimental results demonstrate that MuxTune achieves up to 2.33times higher throughput and 5.29times memory reduction compared to three state-of-the-art baselines.

Emerging Large Language Model (LLM) system patterns, such as disaggregated inference, Mixture-of-Experts (MoE) routing, and asynchronous reinforcement fine-tuning, require flexible point-to-point communication beyond simple collectives. Existing implementations are locked to specific Network Interface Controllers (NICs), hindering integration into inference engines and portability across hardware providers. We present TransferEngine, which bridges the functionality of common NICs to expose a uniform interface. TransferEngine exposes one-sided WriteImm operations with a ImmCounter primitive for completion notification, without ordering assumptions of network transport, transparently managing multiple NICs per GPU. We demonstrate peak throughput of 400 Gbps on both NVIDIA ConnectX-7 and AWS Elastic Fabric Adapter (EFA). We showcase TransferEngine through three production systems: (1) KvCache transfer for disaggregated inference with dynamic scaling, (2) RL weight updates achieving 1.3 seconds for trillion-parameter models, and (3) MoE dispatch/combine implementation exceeding DeepEP decode latency on ConnectX-7, with the first viable latencies on EFA. We demonstrate that our portable point-to-point communication complements collectives while avoiding lock-in.

Large language models (LLMs) require vast amounts of GPU compute to train, but limited availability and high costs of GPUs make homogeneous clusters impractical for many organizations. Instead, assembling heterogeneous clusters by pooling together GPUs of different generations allows them to achieve higher aggregate compute and make use of all available GPUs. However, training on heterogeneous clusters presents several challenges, including load balancing across GPUs, optimizing memory usage to accommodate varying memory capacities, and ensuring communication-efficient training over diverse network interconnects potentially spanning multiple datacenters. In this paper, we make the case that efficient training on heterogeneous clusters requires (1) the integration of pipeline parallelism and data parallelism in a manner that is both communication- and memory-efficient, and (2) a more adaptable configuration of pipeline and data parallelism, which includes the capability to flexibly partition GPUs into asymmetric pipeline parallel stages and to incorporate heterogeneous GPUs within the same data parallelism group. We propose Zorse, the first system to unify all these capabilities while incorporating a planner that automatically configures training strategies for a given workload. Our evaluation shows that Zorse significantly outperforms state-of-the-art systems in heterogeneous training scenarios.

Multi-modal Large Language Models (MLLMs) serving systems commonly employ KV-cache compression to reduce memory footprint. However, existing compression methods introduce significant processing overhead and queuing delays, particularly in concurrent serving scenarios. We present FastCache, a novel serving framework that effectively addresses these challenges through two key innovations: (1) a dynamic batching strategy that optimizes request scheduling across prefill, compression, and decode stages, and (2) an efficient KV-cache memory pool mechanism that eliminates memory fragmentation while maintaining high GPU utilization. Our comprehensive experiments on the GQA and MileBench datasets demonstrate that FastCache achieves up to 19.3times reduction in Time-To-First-Token (TTFT) and 12.1times improvement in throughput compared to state-of-the-art baselines. The system maintains stable performance under high-concurrency scenarios (up to 40 req/s) while reducing average memory consumption by 20\%. These results establish FastCache as an efficient solution for real-world LLM serving systems with KV-cache compression.

Training billion-parameter models requires distributing model states across GPUs using fully sharded data parallel (i.e., ZeRO-3). While ZeRO-3 succeeds on clusters with high-bandwidth NVLink and InfiniBand interconnects, researchers with commodity hardware face severe inter-node all-gather bottlenecks. Existing optimizations take two approaches: GPU memory caching (MiCS, ZeRO++) trades memory capacity for reduced communication, triggering out-of-memory failures on large models; host memory offloading (ZeRO-Offload, ZeRO-Infinity) extends capacity but degrades throughput due to PCIe overhead. We observe that on bandwidth-limited clusters, host memory can serve not as an overflow tier but as a fast caching layer that outperforms inter-node communication. Based on this insight, we propose FCDP, which eliminates redundant inter-node communication while preserving ZeRO-3's minimal GPU memory footprint. FCDP caches forward-pass parameters in host memory and reuses them during the backward pass via fast intra-node all-gather, reducing inter-node all-gather by 50%. For parameter-efficient fine-tuning (PEFT), FCDP selectively communicates only trainable parameters to maximize caching, reducing inter-node traffic by over 99%. In our commodity cluster setup, FCDP achieves up to 100x higher throughput than ZeRO-3 and 51x higher than ZeRO++, while maintaining ZeRO-3's maximum batch size.

This paper addresses emerging system-level challenges in heterogeneous retrieval-augmented generation (RAG) serving, where complex multi-stage workflows and diverse request patterns complicate efficient execution. We present HedraRAG, a runtime system built on a graph-based abstraction that exposes optimization opportunities across stage-level parallelism, intra-request similarity, and inter-request skewness. These opportunities are realized through dynamic graph transformations, such as node splitting, reordering, edge addition, and dependency rewiring, applied to wavefronts of subgraphs spanning concurrent requests. The resulting execution plans are mapped onto hybrid CPU-GPU pipelines to improve resource utilization and reduce latency. Evaluations across a wide range of RAG workflows demonstrate speedups exceeding 1.5x and reaching up to 5x over existing frameworks, showcasing the effectiveness of coordinated generation and retrieval in serving environments.

Quantization is a widely-used compression technology to reduce the overhead of serving large language models (LLMs) on terminal devices and in cloud data centers. However, prevalent quantization methods, such as 8-bit weight-activation or 4-bit weight-only quantization, achieve limited performance improvements due to poor support for low-precision (e.g., 4-bit) activation. This work, for the first time, realizes practical W4A4KV4 serving for LLMs, fully utilizing the INT4 tensor cores on modern GPUs and reducing the memory bottleneck caused by the KV cache. Specifically, we propose a novel fine-grained mixed-precision quantization algorithm (FMPQ) that compresses most activations into 4-bit with negligible accuracy loss. To support mixed-precision matrix multiplication for W4A4 and W4A8, we develop a highly optimized W4Ax kernel. Our approach introduces a novel mixed-precision data layout to facilitate access and fast dequantization for activation and weight tensors, utilizing the GPU's software pipeline to hide the overhead of data loading and conversion. Additionally, we propose fine-grained streaming multiprocessor (SM) scheduling to achieve load balance across different SMs. We integrate the optimized W4Ax kernel into our inference framework, COMET, and provide efficient management to support popular LLMs such as LLaMA-3-70B. Extensive evaluations demonstrate that, when running LLaMA family models on a single A100-80G-SMX4, COMET achieves a kernel-level speedup of 2.88times over cuBLAS and a 2.02 times throughput improvement compared to TensorRT-LLM from an end-to-end framework perspective.

To render each generated token in real-time for users, the Large Language Model (LLM) server generates tokens one by one and streams each token (or group of a few tokens) through the network to the user right after generation, which we refer to as LLM token streaming. However, under unstable network conditions, the LLM token streaming experience could suffer greatly from stalls since one packet loss could block the rendering of later tokens even if the packets containing them arrive on time. With a measurement study, we show that current applications suffer from increased stalls under unstable networks. For this emerging token streaming problem in LLM Chatbots that differs from previous multimedia and text applications, we propose a novel transmission scheme, called Eloquent, which puts newly generated tokens as well as currently unacknowledged tokens in the next outgoing packet. This ensures that each packet contains some new tokens and, in the meantime, is independently rendered when received, avoiding the aforementioned stalls caused by missing packets. Through simulation under various networks, we show Eloquent reduces stall ratio (proportion of token rendering wait time) by 71.0% compared to the retransmission method commonly used by real chatbot applications and by 31.6% compared to the baseline packet duplication scheme. By tailoring Eloquent to fit the token-by-token generation of LLM, we enable the Chatbots to respond like an eloquent speaker for users to better enjoy pervasive AI.

Data sharding, a technique for partitioning and distributing data among multiple servers or nodes, offers enhancements in the scalability, performance, and fault tolerance of extensive distributed systems. Nonetheless, this strategy introduces novel challenges, including load balancing among shards, management of node failures and data loss, and adaptation to evolving data and workload patterns. This paper proposes an innovative approach to tackle these challenges by empowering self-healing nodes with adaptive data sharding. Leveraging concepts such as self-replication, fractal regeneration, sentient data sharding, and symbiotic node clusters, our approach establishes a dynamic and resilient data sharding scheme capable of addressing diverse scenarios and meeting varied requirements. Implementation and evaluation of our approach involve a prototype system simulating a large-scale distributed database across various data sharding scenarios. Comparative analyses against existing data sharding techniques highlight the superior scalability, performance, fault tolerance, and adaptability of our approach. Additionally, the paper delves into potential applications and limitations, providing insights into the future research directions that can further advance this innovative approach.

Prefill-decode (PD) disaggregation has become the standard architecture for large-scale LLM serving, but in practice its deployment boundary is still determined by KVCache transfer. In conventional dense-attention models, prefill generates huge KVCache traffics that keep prefill and decode tightly coupled within a single high-bandwidth network domain, limiting heterogeneous deployment and resource elasticity. Recent hybrid-attention architectures substantially reduce KVCache size, making cross-cluster KVCache transport increasingly plausible. However, smaller KVCache alone does not make heterogeneous cross-datacenter PD serving practical: real workloads remain bursty, request lengths are highly skewed, prefix caches are unevenly distributed, and inter-cluster bandwidth fluctuates. A naive design that fully externalizes prefill can therefore still suffer from congestion, unstable queueing, and poor utilization. We present Prefill-as-a-Service (PrfaaS), a cross-datacenter serving architecture that selectively offloads long-context prefill to standalone, compute-dense prefill clusters and transfers the resulting KVCache over commodity Ethernet to local PD clusters for decode. Rather than treating reduced KVCache as sufficient, PrfaaS combines model-side KV efficiency with system-side selective offloading, bandwidth-aware scheduling, and cache-aware request placement. This design removes the requirement that heterogeneous accelerators share the same low-latency RDMA fabric, enabling independent scaling of prefill and decode capacity across loosely coupled clusters. In a case study using an internal 1T-parameter hybrid model, a PrfaaS-augmented heterogeneous deployment achieves 54% higher serving throughput and 64% lower P90 TTFT than a homogeneous PD baseline, with approximately 15% throughput gain at equal cost, while consuming only modest cross-datacenter bandwidth.

This paper revisits the implementation of Load-balancing Loss (LBL) when training Mixture-of-Experts (MoEs) models. Specifically, LBL for MoEs is defined as N_E sum_{i=1}^{N_E} f_i p_i, where N_E is the total number of experts, f_i represents the frequency of expert i being selected, and p_i denotes the average gating score of the expert i. Existing MoE training frameworks usually employ the parallel training strategy so that f_i and the LBL are calculated within a micro-batch and then averaged across parallel groups. In essence, a micro-batch for training billion-scale LLMs normally contains very few sequences. So, the micro-batch LBL is almost at the sequence level, and the router is pushed to distribute the token evenly within each sequence. Under this strict constraint, even tokens from a domain-specific sequence (e.g., code) are uniformly routed to all experts, thereby inhibiting expert specialization. In this work, we propose calculating LBL using a global-batch to loose this constraint. Because a global-batch contains much more diverse sequences than a micro-batch, which will encourage load balance at the corpus level. Specifically, we introduce an extra communication step to synchronize f_i across micro-batches and then use it to calculate the LBL. Through experiments on training MoEs-based LLMs (up to 42.8B total parameters and 400B tokens), we surprisingly find that the global-batch LBL strategy yields excellent performance gains in both pre-training perplexity and downstream tasks. Our analysis reveals that the global-batch LBL also greatly improves the domain specialization of MoE experts.

Transformer-based Mixture-of-Experts (MoE) models have been driving several recent technological advancements in Natural Language Processing (NLP). These MoE models adopt a router mechanism to determine which experts to activate for routing input tokens. However, existing router mechanisms allocate a fixed number of experts to each token, which neglects the varying importance of different input tokens. In this study, we propose a novel dynamic router mechanism that Dynamically Allocates a variable number of experts for Mixture-of-Experts (DA-MoE) models based on an effective token importance measure. First, we show that the Transformer attention mechanism provides a natural and effective way of calculating token importance. Second, we propose a dynamic router mechanism that effectively decides the optimal number of experts (K) and allocates the top-K experts for each input token. Third, comprehensive experiments on several benchmark datasets demonstrate that our DA-MoE approach consistently outperforms the state-of-the-art Transformer based MoE model on the popular GLUE benchmark.

Small language models are cheap to serve and feasible on local hardware, but strong public 135M-class systems are commonly trained with hundreds of billions to trillions of tokens on large clusters. We study a sharply resource-constrained regime: a complete 134.5M-parameter language-model pipeline executed on one NVIDIA L20 GPU. The released checkpoint, L20-Edu-135M, receives approximately 13B pretraining tokens: 10B FineWeb-Edu tokens followed by a 3B-token educational, mathematics, code, and reasoning mixture. We document the architecture, data gates, cross-source MinHash/LSH near-deduplication, segment deduplication, benchmark-overlap removal, throughput optimization, supervised fine-tuning (SFT) with weight interpolation, and reinforcement learning from verifiable rewards (RLVR) on GSM8K. In a self-run zero-shot six-task harness, L20-Edu-135M obtains a mean score of 0.4150. It trails SmolLM-135M (0.4767) and SmolLM2-135M (0.4917), but its mean is 87.1% of SmolLM-135M's while its nominal token count is 2.17% as large. This ratio is descriptive, not evidence of statistical equivalence or a controlled scaling law. The model exceeds several older 100M-160M public baselines under the same harness. Direct GRPO-style RLVR decreases GSM8K exact-match accuracy from 1.82% to 1.59% (192-token completions) and 1.21% (320-token completions). These single-run results identify a concrete failure mode rather than establishing a general lower bound on RLVR. The contribution is an auditable resource-constrained case study, not a state-of-the-art claim.

Fine-grained sparsity promises higher parametric capacity without proportional per-token compute, but often suffers from training instability, load balancing, and communication overhead. We introduce STEM (Scaling Transformers with Embedding Modules), a static, token-indexed approach that replaces the FFN up-projection with a layer-local embedding lookup while keeping the gate and down-projection dense. This removes runtime routing, enables CPU offload with asynchronous prefetch, and decouples capacity from both per-token FLOPs and cross-device communication. Empirically, STEM trains stably despite extreme sparsity. It improves downstream performance over dense baselines while reducing per-token FLOPs and parameter accesses (eliminating roughly one-third of FFN parameters). STEM learns embedding spaces with large angular spread which enhances its knowledge storage capacity. More interestingly, this enhanced knowledge capacity comes with better interpretability. The token-indexed nature of STEM embeddings allows simple ways to perform knowledge editing and knowledge injection in an interpretable manner without any intervention in the input text or additional computation. In addition, STEM strengthens long-context performance: as sequence length grows, more distinct parameters are activated, yielding practical test-time capacity scaling. Across 350M and 1B model scales, STEM delivers up to ~3--4% accuracy improvements overall, with notable gains on knowledge and reasoning-heavy benchmarks (ARC-Challenge, OpenBookQA, GSM8K, MMLU). Overall, STEM is an effective way of scaling parametric memory while providing better interpretability, better training stability and improved efficiency.

Hardware prefetching is critical to fill the performance gap between CPU speeds and slower memory accesses. With multicore architectures becoming commonplace, traditional prefetchers are severely challenged. Independent core operation creates significant redundancy (up to 20% of prefetch requests are duplicates), causing unnecessary memory bus traffic and wasted bandwidth. Furthermore, cutting-edge prefetchers such as Pythia suffer from about a 10% performance loss when scaling from a single-core to a four-core system. To solve these problems, we propose CRL-Pythia, a coordinated reinforcement learning based prefetcher specifically designed for multicore systems. In this work, CRL-Pythia addresses these issues by enabling cross-core sharing of information and cooperative prefetching decisions, which greatly reduces redundant prefetch requests and improves learning convergence across cores. Our experiments demonstrate that CRL-Pythia outperforms single Pythia configurations in all cases, with approximately 12% IPC (instructions per cycle) improvement for bandwidth-constrained workloads, while imposing moderate hardware overhead. Our sensitivity analyses also verify its robustness and scalability, thereby making CRL-Pythia a practical and efficient solution to contemporary multicore systems.

Many deep learning applications benefit from using large models with billions of parameters. Training these models is notoriously expensive due to the need for specialized HPC clusters. In this work, we consider alternative setups for training large models: using cheap "preemptible" instances or pooling existing resources from multiple regions. We analyze the performance of existing model-parallel algorithms in these conditions and find configurations where training larger models becomes less communication-intensive. Based on these findings, we propose SWARM parallelism, a model-parallel training algorithm designed for poorly connected, heterogeneous and unreliable devices. SWARM creates temporary randomized pipelines between nodes that are rebalanced in case of failure. We empirically validate our findings and compare SWARM parallelism with existing large-scale training approaches. Finally, we combine our insights with compression strategies to train a large Transformer language model with 1B shared parameters (approximately 13B before sharing) on preemptible T4 GPUs with less than 200Mb/s network.

Large Language Model (LLM) workloads have distinct prefill and decode phases with different compute and memory requirements which should ideally be accounted for when scheduling input queries across different LLM instances in a cluster. However existing scheduling algorithms treat LLM workloads as monolithic jobs without considering the distinct characteristics of the two phases in each workload. This leads to sub-optimal scheduling and increased response latency. In this work, we propose a heuristic-guided reinforcement learning-based intelligent router for data-driven and workload-aware scheduling. Our router leverages a trainable response-length predictor, and a novel formulation for estimating the impact of mixing different workloads to schedule queries across LLM instances and achieve over 11\% lower end-to-end latency than existing approaches.

LLMs have traditionally scaled along dense dimensions, where performance is coupled with near-linear increases in computational cost. While MoE decouples capacity from compute, it introduces large memory overhead and hardware efficiency challenges. To overcome these, we propose token-indexed parameters as a novel, orthogonal scaling axis that decouple model capacity from FLOPs. Specifically, we introduce Joint-Token (JTok) and Mixture of Joint-Token (JTok-M), which augment Transformer layers with modulation vectors retrieved from auxiliary embedding tables. These vectors modulate the backbone via lightweight, element-wise operations, incurring negligible FLOPs overhead. Extensive experiments on both dense and MoE backbones, spanning from 650M (190M + 460M embedding) to 61B (17B + 44B embedding) total parameters, demonstrate that our approach consistently reduces validation loss and significantly improves downstream task performance (e.g., +4.1 on MMLU, +8.3 on ARC, +8.9 on CEval). Rigorous isoFLOPs analysis further confirms that JTok-M fundamentally shifts the quality-compute Pareto frontier, achieving comparable model quality with 35% less compute relative to vanilla MoE architectures, and we validate that token-indexed parameters exhibit a predictable power-law scaling behavior. Moreover, our efficient implementation ensures that the overhead introduced by JTok and JTok-M remains marginal.

Over the past year, the vLLM Semantic Router project has released a series of work spanning: (1) core routing mechanisms -- signal-driven routing, context-length pool routing, router performance engineering, policy conflict detection, low-latency embedding models, category-aware semantic caching, user-feedback-driven routing adaptation, hallucination detection, and hierarchical content-safety classification for privacy and jailbreak protection; (2) fleet optimization -- fleet provisioning and energy-efficiency analysis; (3) agentic and multimodal routing -- multimodal agent routing, tool selection, CUA security, and multi-turn context memory and safety; (4) governance and standards -- inference routing protocols and multi-provider API extensions. Each paper tackled a specific problem in LLM inference, but the problems are not independent; for example, fleet provisioning depends on the routing policy, which depends on the workload mix, shifting as organizations adopt agentic and multimodal workloads. This paper distills those results into the Workload-Router-Pool (WRP) architecture, a three-dimensional framework for LLM inference optimization. Workload characterizes what the fleet serves (chat vs. agent, single-turn vs. multi-turn, warm vs. cold, prefill-heavy vs. decode-heavy). Router determines how each request is dispatched (static semantic rules, online bandit adaptation, RL-based model selection, quality-aware cascading). Pool defines where inference runs (homogeneous vs. heterogeneous GPU, disaggregated prefill/decode, KV-cache topology). We map our prior work onto a 3x3 WRP interaction matrix, identify which cells we have covered and which remain open, and propose twenty-one concrete research directions at the intersections, each grounded in our prior measurements, tiered by maturity from engineering-ready to open research.

The most popular framework for distributed training of machine learning models is the (synchronous) parameter server (PS). This paradigm consists of n workers, which iteratively compute updates of the model parameters, and a stateful PS, which waits and aggregates all updates to generate a new estimate of model parameters and sends it back to the workers for a new iteration. Transient computation slowdowns or transmission delays can intolerably lengthen the time of each iteration. An efficient way to mitigate this problem is to let the PS wait only for the fastest n-b updates, before generating the new parameters. The slowest b workers are called backup workers. The optimal number b of backup workers depends on the cluster configuration and workload, but also (as we show in this paper) on the hyper-parameters of the learning algorithm and the current stage of the training. We propose DBW, an algorithm that dynamically decides the number of backup workers during the training process to maximize the convergence speed at each iteration. Our experiments show that DBW 1) removes the necessity to tune b by preliminary time-consuming experiments, and 2) makes the training up to a factor 3 faster than the optimal static configuration.

Large-scale distributed training in production datacenters constitutes a challenging workload bottlenecked by network communication. In response, both major industry players (e.g., Ultra Ethernet Consortium) and parts of academia have surprisingly, and almost unanimously, agreed that packet spraying is necessary to improve the performance of large-scale distributed training workloads. In this paper, we challenge this prevailing belief and pose the question: How close can a singlepath transport approach an optimal multipath transport? We demonstrate that singlepath transport (from a NIC's perspective) is sufficient and can perform nearly as well as an ideal multipath transport with packet spraying, particularly in the context of distributed training in leaf-spine topologies. Our assertion is based on four key observations about workloads driven by collective communication patterns: (i) flows within a collective start almost simultaneously, (ii) flow sizes are nearly equal, (iii) the completion time of a collective is more crucial than individual flow completion times, and (iv) flows can be split upon arrival. We analytically prove that singlepath transport, using minimal flow splitting (at the application layer), is equivalent to an ideal multipath transport with packet spraying in terms of maximum congestion. Our preliminary evaluations support our claims. This paper suggests an alternative agenda for developing next-generation transport protocols tailored for large-scale distributed training.

We investigate a failure mode of large language models (LLMs) in which plain-text prompts elicit excessive outputs, a phenomenon we term Overflow. Unlike jailbreaks or prompt injection, Overflow arises under ordinary interaction settings and can lead to elevated serving cost, latency, and cross-user performance degradation, particularly when scaled across many requests. Beyond usability, the stakes are economic and environmental: unnecessary tokens increase per-request cost and energy consumption, compounding into substantial operational spend and carbon footprint at scale. Moreover, Overflow represents a practical vector for compute amplification and service degradation in shared environments. We introduce BenchOverflow, a model-agnostic benchmark of nine plain-text prompting strategies that amplify output volume without adversarial suffixes or policy circumvention. Using a standardized protocol with a fixed budget of 5000 new tokens, we evaluate nine open- and closed-source models and observe pronounced rightward shifts and heavy tails in length distributions. Cap-saturation rates (CSR@1k/3k/5k) and empirical cumulative distribution functions (ECDFs) quantify tail risk; within-prompt variance and cross-model correlations show that Overflow is broadly reproducible yet heterogeneous across families and attack vectors. A lightweight mitigation-a fixed conciseness reminder-attenuates right tails and lowers CSR for all strategies across the majority of models. Our findings position length control as a measurable reliability, cost, and sustainability concern rather than a stylistic quirk. By enabling standardized comparison of length-control robustness across models, BenchOverflow provides a practical basis for selecting deployments that minimize resource waste and operating expense, and for evaluating defenses that curb compute amplification without eroding task performance.

Multi-agent applications often execute complex tasks as multi-stage workflows, where each stage is an LLM call whose output becomes part of context for subsequent steps. Existing LLM serving systems largely assume homogeneous clusters with identical model replicas. This design overlooks the potential of heterogeneous deployments, where models of different sizes and capabilities enable finer trade-offs between latency and performance. However, heterogeneity introduces new challenges in scheduling across models with diverse throughput and performance. We present Chimera, a predictive scheduling system for multi-agent workflow serving on heterogeneous LLM clusters that jointly improves end-to-end latency and task performance. Chimera applies semantic routing to estimate per-model confidence scores for each request, predicts the total remaining output length of the workflow, and estimates per-model congestion using in-flight predicted token volumes for load balancing. We evaluate Chimera on representative agentic workflows for code generation and math reasoning using multiple heterogeneous LLM configurations. Across comparable settings, Chimera traces the best latency-performance frontier, reducing end-to-end latency by 1.2--2.4times and improving task performance by 8.0-9.5 percentage points on average over competitive baselines including vLLM.

Retrieval-Augmented Generation (RAG) systems critically depend on effective document chunking strategies to balance retrieval quality, latency, and operational cost. Traditional chunking approaches, such as fixed-size, rule-based, or fully agentic chunking, often suffer from high token consumption, redundant text generation, limited scalability, and poor debuggability, especially for large-scale web content ingestion. In this paper, we propose Web Retrieval-Aware Chunking (W-RAC), a novel, cost-efficient chunking framework designed specifically for web-based documents. W-RAC decouples text extraction from semantic chunk planning by representing parsed web content as structured, ID-addressable units and leveraging large language models (LLMs) only for retrieval-aware grouping decisions rather than text generation. This significantly reduces token usage, eliminates hallucination risks, and improves system observability.Experimental analysis and architectural comparison demonstrate that W-RAC achieves comparable or better retrieval performance than traditional chunking approaches while reducing chunking-related LLM costs by an order of magnitude.

Training large language models requires distributing computation across many accelerators, yet practitioners select parallelism strategies (data, tensor, pipeline, ZeRO) through trial and error because no unified systematic framework predicts their behavior. We introduce placement semantics: each strategy is specified by how it places four training states (parameters, optimizer, gradients, activations) across devices using five modes (replicated, sharded, sharded-with-gather, materialized, offloaded). From placement alone, without implementation details, we derive memory consumption and communication volume. Our predictions match published results exactly: ZeRO-3 uses 8x less memory than data parallelism at 1.5x communication cost, as reported in the original paper. We prove two conditions (gradient integrity, state consistency) are necessary and sufficient for distributed training to match single-device results, and provide composition rules for combining strategies safely. The framework unifies ZeRO Stages 1-3, Fully Sharded Data Parallel (FSDP), tensor parallelism, and pipeline parallelism as instances with different placement choices.

Large Language Model or LLM inference has two phases, the prompt (or prefill) phase to output the first token and the extension (or decoding) phase to the generate subsequent tokens. In this work, we propose an efficient parallelization scheme, KV-Runahead to accelerate the prompt phase. The key observation is that the extension phase generates tokens faster than the prompt phase because of key-value cache (KV-cache). Hence, KV-Runahead parallelizes the prompt phase by orchestrating multiple processes to populate the KV-cache and minimizes the time-to-first-token (TTFT). Dual-purposing the KV-cache scheme has two main benefits. Fist, since KV-cache is designed to leverage the causal attention map, we minimize computation and computation automatically. Second, since it already exists for the exten- sion phase, KV-Runahead is easy to implement. We further propose context-level load-balancing to handle uneven KV-cache generation (due to the causal attention) and to optimize TTFT. Compared with an existing parallelization scheme such as tensor or sequential parallelization where keys and values are locally generated and exchanged via all-gather collectives, our experimental results demonstrate that KV-Runahead can offer over 1.4x and 1.6x speedups for Llama 7B and Falcon 7B respectively.

This paper challenges the well-established paradigm for building any-to-any networks for training Large Language Models (LLMs). We show that LLMs exhibit a unique communication pattern where only small groups of GPUs require high-bandwidth any-to-any communication within them, to achieve near-optimal training performance. Across these groups of GPUs, the communication is insignificant, sparse, and homogeneous. We propose a new network architecture that closely resembles the communication requirement of LLMs. Our architecture partitions the cluster into sets of GPUs interconnected with non-blocking any-to-any high-bandwidth interconnects that we call HB domains. Across the HB domains, the network only connects GPUs with communication demands. We call this network a "rail-only" connection, and show that our proposed architecture reduces the network cost by up to 75% compared to the state-of-the-art any-to-any Clos networks without compromising the performance of LLM training.

Inference serving for large language models (LLMs) is the key to unleashing their potential in people's daily lives. However, efficient LLM serving remains challenging today because the requests are inherently heterogeneous and unpredictable in terms of resource and latency requirements, as a result of the diverse applications and the dynamic execution nature of LLMs. Existing systems are fundamentally limited in handling these characteristics and cause problems such as severe queuing delays, poor tail latencies, and SLO violations. We introduce Llumnix, an LLM serving system that reacts to such heterogeneous and unpredictable requests by runtime rescheduling across multiple model instances. Similar to context switching across CPU cores in modern operating systems, Llumnix reschedules requests to improve load balancing and isolation, mitigate resource fragmentation, and differentiate request priorities and SLOs. Llumnix implements the rescheduling with an efficient and scalable live migration mechanism for requests and their in-memory states, and exploits it in a dynamic scheduling policy that unifies the multiple rescheduling scenarios elegantly. Our evaluations show that Llumnix improves tail latencies by an order of magnitude, accelerates high-priority requests by up to 1.5x, and delivers up to 36% cost savings while achieving similar tail latencies, compared against state-of-the-art LLM serving systems. Llumnix is publicly available at https://github.com/AlibabaPAI/llumnix.

Large language models (LLMs) such as GPT-3, OPT, and LLaMA have demonstrated remarkable accuracy in a wide range of tasks. However, training these models can incur significant expenses, often requiring tens of thousands of GPUs for months of continuous operation. Typically, this training is carried out in specialized GPU clusters equipped with homogeneous high-speed Remote Direct Memory Access (RDMA) network interface cards (NICs). The acquisition and maintenance of such dedicated clusters is challenging. Current LLM training frameworks, like Megatron-LM and Megatron-DeepSpeed, focus primarily on optimizing training within homogeneous cluster settings. In this paper, we introduce Holmes, a training framework for LLMs that employs thoughtfully crafted data and model parallelism strategies over the heterogeneous NIC environment. Our primary technical contribution lies in a novel scheduling method that intelligently allocates distinct computational tasklets in LLM training to specific groups of GPU devices based on the characteristics of their connected NICs. Furthermore, our proposed framework, utilizing pipeline parallel techniques, demonstrates scalability to multiple GPU clusters, even in scenarios without high-speed interconnects between nodes in distinct clusters. We conducted comprehensive experiments that involved various scenarios in the heterogeneous NIC environment. In most cases, our framework achieves performance levels close to those achievable with homogeneous RDMA-capable networks (InfiniBand or RoCE), significantly exceeding training efficiency within the pure Ethernet environment. Additionally, we verified that our framework outperforms other mainstream LLM frameworks under heterogeneous NIC environment in terms of training efficiency and can be seamlessly integrated with them.

This document presents a vision for a novel AI infrastructure design that has been initially validated through inference simulations on state-of-the-art large language models. Advancements in deep learning and specialized hardware have driven the rapid growth of large language models (LLMs) and generative AI systems. However, traditional GPU-centric architectures face scalability challenges for inference workloads due to limitations in memory capacity, bandwidth, and interconnect scaling. To address these issues, the FengHuang Platform, a disaggregated AI infrastructure platform, is proposed to overcome memory and communication scaling limits for AI inference. FengHuang features a multi-tier shared-memory architecture combining high-speed local memory with centralized disaggregated remote memory, enhanced by active tensor paging and near-memory compute for tensor operations. Simulations demonstrate that FengHuang achieves up to 93% local memory capacity reduction, 50% GPU compute savings, and 16x to 70x faster inter-GPU communication compared to conventional GPU scaling. Across workloads such as GPT-3, Grok-1, and QWEN3-235B, FengHuang enables up to 50% GPU reductions while maintaining end-user performance, offering a scalable, flexible, and cost-effective solution for AI inference infrastructure. FengHuang provides an optimal balance as a rack-level AI infrastructure scale-up solution. Its open, heterogeneous design eliminates vendor lock-in and enhances supply chain flexibility, enabling significant infrastructure and power cost reductions.

Large Language Model (LLM) serving systems remain fundamentally fragile, where frequent hardware faults in hyperscale clusters trigger disproportionate service outages in the software stack. Current recovery mechanisms are prohibitively slow, often requiring up to 10 minutes to reinitialize resources and reload massive model weights. We introduce KevlarFlow, a fault tolerant serving architecture designed to bridge the gap between hardware unreliability and service availability. KevlarFlow leverages 1) decoupled model parallelism initialization, 2) dynamic traffic rerouting, and 3) background KV cache replication to maintain high throughput during partial failures. Our evaluation demonstrates that KevlarFlow reduces mean-time-to-recovery (MTTR) by 20x and, under failure conditions, improves average latency by 3.1x, 99th percentile (p99) latency by 2.8x, average time-to-first-token (TTFT) by 378.9x, and p99 TTFT by 574.6x with negligible runtime overhead in comparison to state-of-the-art LLM serving systems.

We survey the large language model (LLM) serving area to understand the intricate dynamics between cost-efficiency and accuracy, which is magnified by the growing need for longer contextual understanding when deploying models at a massive scale. Our findings reveal that works in this space optimize along three distinct but conflicting goals: improving serving context length (C), improving serving accuracy (A), and improving serving performance (P). Drawing inspiration from the CAP theorem in databases, we propose a CAP principle for LLM serving, which suggests that any optimization can improve at most two of these three goals simultaneously. Our survey categorizes existing works within this framework. We find the definition and continuity of user-perceived measurement metrics are crucial in determining whether a goal has been met, akin to prior CAP databases in the wild. We recognize the CAP principle for LLM serving as a guiding principle, rather than a formal theorem, to inform designers of the inherent and dynamic trade-offs in serving models. As serving accuracy and performance have been extensively studied, this survey focuses on works that extend serving context length and address the resulting challenges.

This paper proposes a novel admission and routing scheme which takes into account arbitrarily assigned priorities for network flows. The presented approach leverages the centralized Software Defined Networking (SDN) capabilities in order to do so. Exact and heuristic approaches to the stated Priority Flow Admission and Routing (PFAR) problem are provided. The exact approach which provides an optimal solution is based on Integer Linear Programming (ILP). Given the potentially long running time required to find an exact and optimal solution, a heuristic approach is proposed; this approach is based on Genetic Algorithms (GAs). In order to effectively estimate the performance of the proposed approaches, a simulator that is capable of generating semi-random network topologies and flows has been developed. Experimental results for large problem instances (up 50 network nodes and thousands of network flows), show that: i) an optimal solution can be often found in few seconds (even milliseconds), and ii) the heuristic approach yields close-to-optimal solutions (approximately 95\% of the optimal) in a fixed amount of time; these experimental results demonstrate the pertinence of the proposed approaches.

Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to Autoregressive Models (ARMs), utilizing parallel decoding to overcome sequential bottlenecks. However, existing research focuses primarily on kernel-level optimizations, lacking a holistic serving framework that addresses the unique memory dynamics of diffusion processes in production. We identify a critical "memory footprint crisis" specific to dLLMs, driven by monolithic logit tensors and the severe resource oscillation between compute-bound "Refresh" phases and bandwidth-bound "Reuse" phases. To bridge this gap, we present dLLM-Serve, an efficient dLLM serving system that co-optimizes memory footprint, computational scheduling, and generation quality. dLLM-Serve introduces Logit-Aware Activation Budgeting to decompose transient tensor peaks, a Phase-Multiplexed Scheduler to interleave heterogeneous request phases, and Head-Centric Sparse Attention to decouple logical sparsity from physical storage. We evaluate dLLM-Serve on diverse workloads (LiveBench, Burst, OSC) and GPUs (RTX 4090, L40S). Relative to the state-of-the-art baseline, dLLM-Serve improves throughput by 1.61times-1.81times on the consumer-grade RTX 4090 and 1.60times-1.74times on the server-grade NVIDIA L40S, while reducing tail latency by nearly 4times under heavy contention. dLLM-Serve establishes the first blueprint for scalable dLLM inference, converting theoretical algorithmic sparsity into tangible wall-clock acceleration across heterogeneous hardware.

The increasing usage of Large Language Models (LLMs) has resulted in a surging demand for planet-scale serving systems, where tens of thousands of GPUs continuously serve hundreds of millions of users. Consequently, throughput (under reasonable latency constraints) has emerged as a key metric that determines serving systems' performance. To boost throughput, various methods of inter-device parallelism (e.g., data, tensor, pipeline) have been explored. However, existing methods do not consider overlapping the utilization of different resources within a single device, leading to underutilization and sub-optimal performance. We propose NanoFlow, a novel serving framework that exploits intra-device parallelism, which overlaps the usage of resources including compute, memory, and network within a single device through operation co-scheduling. To exploit intra-device parallelism, NanoFlow introduces two key innovations: First, NanoFlow splits requests into nano-batches at the granularity of operations, which breaks the dependency of sequential operations in LLM inference and enables overlapping; then, to get benefit from overlapping, NanoFlow uses an operation-level pipeline with execution unit scheduling, which partitions the device's functional units and simultaneously executes different operations in each unit. NanoFlow automates the pipeline setup using a parameter search algorithm, which enables easily porting NanoFlow to different models. We implement NanoFlow on NVIDIA GPUs and evaluate end-to-end serving throughput on several popular models such as LLaMA-2-70B, Mixtral 8x7B, LLaMA-3-8B, etc.. With practical workloads, NanoFlow provides 1.91x throughput boost compared to state-of-the-art serving systems achieving 59% to 72% of optimal throughput across ported models.

Tokenization is the first -- and often least scrutinized -- step of most NLP pipelines. Standard algorithms for learning tokenizers rely on frequency-based objectives, which favor languages dominant in the training data and consequently leave lower-resource languages with tokenizations that are disproportionately longer, morphologically implausible, or even riddled with <UNK> placeholders. This phenomenon ultimately amplifies computational and financial inequalities between users from different language backgrounds. To remedy this, we introduce Parity-aware Byte Pair Encoding (BPE), a variant of the widely-used BPE algorithm. At every merge step, Parity-aware BPE maximizes the compression gain of the currently worst-compressed language, trading a small amount of global compression for cross-lingual parity. We find empirically that Parity-aware BPE leads to more equitable token counts across languages, with negligible impact on global compression rate and no substantial effect on language-model performance in downstream tasks.