doc_id stringlengths 11 42 | source_type stringclasses 5
values | source_url stringlengths 32 176 | license stringclasses 8
values | title stringlengths 3 654 | content stringlengths 439 232k | project stringclasses 35
values | tags listlengths 2 31 | word_count int32 71 33.5k | lang stringclasses 1
value | quality_score float64 0.35 0.9 | scraped_at timestamp[us, tz=UTC]date 2026-09-21 04:26:58 2026-09-21 12:13:36 |
|---|---|---|---|---|---|---|---|---|---|---|---|
arxiv-1907.12817-11 | arxiv | https://arxiv.org/abs/1907.12817 | CC-BY-4.0 | Increasing Scalability of Process Mining using Event Dataframes: How Data Structure Matters — 7 Conclusion | Title: Increasing Scalability of Process Mining using Event Dataframes: How Data Structure Matters — 7 Conclusion
arXiv: 1907.12817
## 7 Conclusion
In this paper, a combination of columnar storages (like Apache Parquet and Apache ORC) and dataframes (popular implementations are Pandas and Apache Spark) is used in ord... | papers | [
"paper",
"arxiv_id:1907.12817",
"relevance:4",
"chunk:11"
] | 520 | en | 0.5 | 2026-09-21T09:48:30.698048Z |
arxiv-2002.08102-00 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake
arXiv: 2002.08102 | License: CC-BY-4.0
Abstract: The optimization of query execution plans is known to be crucial for reducing the query execution time. In particular, query optimization has been studied thoroughly for relational databases... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:0"
] | 438 | en | 0.45 | 2026-09-21T09:47:35.173489Z |
arxiv-2002.08102-01 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake — 1 Introduction | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake — 1 Introduction
arXiv: 2002.08102
## 1 Introduction
Advances in the technologies for data generation and ingestion facilitate the collection of large volumes of data from where valuable knowledge can be extracted. However, the wide varie... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:1"
] | 391 | en | 0.4 | 2026-09-21T09:47:35.173499Z |
arxiv-2002.08102-02 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake — 2 Our Approach | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake — 2 Our Approach
arXiv: 2002.08102
## 2 Our Approach
### 2.1 Preliminaries
In this section, we present basic concepts required to understand this work. SPARQL queries can be partitioned into groups of acyclic patterns that share exactly ... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:2"
] | 801 | en | 0.45 | 2026-09-21T09:47:35.173504Z |
arxiv-2002.08102-03 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake — 3 Experiment | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake — 3 Experiment
arXiv: 2002.08102
## 3 Experiment
We empirically study two different kinds of query execution plans in order to evaluate the proposed heuristics. The QEPs are as follows:
a)*Physical-Design-Unaware QEP*: A QEP not respecti... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:3"
] | 822 | en | 0.425 | 2026-09-21T09:47:35.173508Z |
arxiv-2002.08102-04 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake — 4 Related Work | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake — 4 Related Work
arXiv: 2002.08102
## 4 Related Work
#### Federated Databases.
The problem of integrating data from dissimilar data sources has been addressed in the literature by implementing the mediator and wrapper architecture propos... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:4"
] | 533 | en | 0.5 | 2026-09-21T09:47:35.173512Z |
arxiv-2002.08102-05 | arxiv | https://arxiv.org/abs/2002.08102 | CC-BY-4.0 | Optimizing Federated Queries Based on the Physical Design of a Data Lake — 5 Conclusion | Title: Optimizing Federated Queries Based on the Physical Design of a Data Lake — 5 Conclusion
arXiv: 2002.08102
## 5 Conclusion
In this paper we present two rather simple heuristics that aim at improving the query performance by considering the physical design of the Data Lake compared to state-of-the-art query exec... | papers | [
"paper",
"arxiv_id:2002.08102",
"relevance:5",
"chunk:5"
] | 206 | en | 0.45 | 2026-09-21T09:47:35.173515Z |
arxiv-2010.12597-00 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework | Title: DBLog: A Watermark Based Change-Data-Capture Framework
arXiv: 2010.12597 | License: CC-BY-4.0
Abstract: It is a commonly observed pattern for applications to utilize multiple heterogeneous databases where each is used to serve a specific need such as storing the canonical form of data or providing advanced sear... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:0"
] | 1,383 | en | 0.65 | 2026-09-21T09:48:27.523371Z |
arxiv-2010.12597-01 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 2. Related Work | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 2. Related Work
arXiv: 2010.12597
## 2. Related Work
We evaluated a series of existing offerings such as: Databus (Das et al., 2012), Debezium (Debezium documentation, 2020), Maxwell (Zendesk, 2014), MySQLStreamer (Prem Santosh Udaya Shankar, 2016), Spin... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:1"
] | 825 | en | 0.6 | 2026-09-21T09:48:27.523383Z |
arxiv-2010.12597-02 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog
arXiv: 2010.12597
## 3. DBLog
DBLog is a Java-based framework, able to capture changed rows from a database's transaction log and to capture the full state of a database by executing selects on tables. Selects are executed in chunks and are inte... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:2"
] | 328 | en | 0.575 | 2026-09-21T09:48:27.523387Z |
arxiv-2010.12597-03 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.1. Transaction log capture | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.1. Transaction log capture
arXiv: 2010.12597
### 3.1. Transaction log capture
The framework requires a database to emit an event for each changed row in commit order. In MySQL and PostgreSQL a replication protocol exists where the database d... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:3"
] | 257 | en | 0.525 | 2026-09-21T09:48:27.523391Z |
arxiv-2010.12597-04 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.2. Full state capture | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.2. Full state capture
arXiv: 2010.12597
### 3.2. Full state capture
As transaction logs typically have limited retention they can not be used to reconstruct the full source dataset. When attempting to solve this problem, two of the major cha... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:4"
] | 1,213 | en | 0.525 | 2026-09-21T09:48:27.523395Z |
arxiv-2010.12597-05 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.3. Database Support | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 3. DBLog › 3.3. Database Support
arXiv: 2010.12597
### 3.3. Database Support
In order to use DBLog a database needs to emit changed rows from a linear history in commit order and support non-stale reads. These conditions are fulfilled by systems like MyS... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:5"
] | 268 | en | 0.55 | 2026-09-21T09:48:27.523398Z |
arxiv-2010.12597-06 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 4. DBLog in Production | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 4. DBLog in Production
arXiv: 2010.12597
## 4. DBLog in Production
DBLog is the foundation of the MySQL and PostgreSQL Connectors at Netflix. They are both used in our data synchronization and enrichment platform called Delta (Andreakis et al., 2019). DB... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:6"
] | 424 | en | 0.55 | 2026-09-21T09:48:27.523402Z |
arxiv-2010.12597-07 | arxiv | https://arxiv.org/abs/2010.12597 | CC-BY-4.0 | DBLog: A Watermark Based Change-Data-Capture Framework — 5. Conclusions | Title: DBLog: A Watermark Based Change-Data-Capture Framework — 5. Conclusions
arXiv: 2010.12597
## 5. Conclusions
In this paper, we presented a novel watermark based CDC framework. DBLog capabilities extend the capture of changed rows in real-time from a database transaction log to also extract the full state of a d... | papers | [
"paper",
"arxiv_id:2010.12597",
"relevance:4",
"chunk:7"
] | 323 | en | 0.525 | 2026-09-21T09:48:27.523405Z |
arxiv-2103.10169-00 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile
arXiv: 2103.10169 | License: CC-BY-4.0
Abstract: Jet is an open-source, high-performance, distributed stream processor built at Hazelcast during the last five years. Jet was engineered with millisecond latency on the 99.99th percentile as it... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:0"
] | 1,407 | en | 0.575 | 2026-09-21T09:48:21.601943Z |
arxiv-2103.10169-01 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 2. System Overview | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 2. System Overview
arXiv: 2103.10169
## 2. System Overview
Jet adopts the standard streaming data flow model, in which computations form a directed acyclic graph of operators (vertices) that apply transformations on data streams (edges). ... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:1"
] | 758 | en | 0.55 | 2026-09-21T09:48:21.601954Z |
arxiv-2103.10169-02 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 3. Execution Model | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 3. Execution Model
arXiv: 2103.10169
## 3. Execution Model
Jet's execution model involves the deployment of stateful dataflow graphs that we explain in Section 3.1 and the execution of computations (Section 3.2). Section 3.3 clarifies how... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:2"
] | 1,292 | en | 0.475 | 2026-09-21T09:48:21.601958Z |
arxiv-2103.10169-05 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.2. In-Memory Data Grid (IMDG) | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.2. In-Memory Data Grid (IMDG)
arXiv: 2103.10169
### 4.2. In-Memory Data Grid (IMDG)
One key feature that distinguishes Jet from other modern scale-out streaming systems is that, although fault-tolerant and highly a... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:5"
] | 337 | en | 0.425 | 2026-09-21T09:48:21.601969Z |
arxiv-2103.10169-07 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.4. Fault Tolerance and Processing Guarantees | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.4. Fault Tolerance and Processing Guarantees
arXiv: 2103.10169
### 4.4. Fault Tolerance and Processing Guarantees
Jet's approach to fault tolerance is based on the seminal work on distributed snapshots by Chandy an... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:7"
] | 311 | en | 0.475 | 2026-09-21T09:48:21.601976Z |
arxiv-2103.10169-08 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.5. Assumptions and External Systems | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.5. Assumptions and External Systems
arXiv: 2103.10169
### 4.5. Assumptions and External Systems
Jet and most streaming systems (Armbrust et al., 2018; Carbone et al., 2015; Noghabi et al., 2017; Jacques-Silva et al... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:8"
] | 417 | en | 0.6 | 2026-09-21T09:48:21.601979Z |
arxiv-2103.10169-09 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.6. Fault Tolerance via Active Replication | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 4. State Management › 4.6. Fault Tolerance via Active Replication
arXiv: 2103.10169
### 4.6. Fault Tolerance via Active Replication
From our experience with clients and Jet users we have noticed a very interesting trade-off: by implementi... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:9"
] | 257 | en | 0.5 | 2026-09-21T09:48:21.601982Z |
arxiv-2103.10169-10 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 5. Millisecond Latency on the JVM | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 5. Millisecond Latency on the JVM
arXiv: 2103.10169
## 5. Millisecond Latency on the JVM
The key competence of Jet is its capacity to perform ultra low-latency processing. Four design decisions drive this achievement. First, Jet schedules... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:10"
] | 463 | en | 0.4 | 2026-09-21T09:48:21.601986Z |
arxiv-2103.10169-11 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 6. Use Cases | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 6. Use Cases
arXiv: 2103.10169
## 6. Use Cases
Hazelcast clients and open-source users deploy Jet in many different ways; many of which we found surprising. Here we report on a subset of them that we believe are closest to the interest of... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:11"
] | 915 | en | 0.6 | 2026-09-21T09:48:21.601989Z |
arxiv-2103.10169-13 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.1. Experimental Methodology | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.1. Experimental Methodology
arXiv: 2103.10169
### 7.1. Experimental Methodology
We evaluate Jet v4.3 in an Amazon Cloud cluster comprising 1, 5 and 10 c5.4xlarge machines with 16 vCPUs and 32 GB RAM each. Jet runs on Or... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:13"
] | 590 | en | 0.5 | 2026-09-21T09:48:21.601995Z |
arxiv-2103.10169-15 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.3. Throughput vs. Latency: 10ms Slide on Q5 | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.3. Throughput vs. Latency: 10ms Slide on Q5
arXiv: 2103.10169
### 7.3. Throughput vs. Latency: 10ms Slide on Q5
In this experiment we want to push Jet to its limit with respect to throughput per CPU-core on a windowed a... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:15"
] | 257 | en | 0.4 | 2026-09-21T09:48:21.602001Z |
arxiv-2103.10169-17 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.5. Latency: Windowed Aggregates & Joins | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 7. Experiments › 7.5. Latency: Windowed Aggregates & Joins
arXiv: 2103.10169
### 7.5. Latency: Windowed Aggregates & Joins
In this experiment we want to show the capacity of Jet to compute very low-latency aggregates and windowed joins. T... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:17"
] | 206 | en | 0.4 | 2026-09-21T09:48:21.602008Z |
arxiv-2103.10169-20 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 8. Related Work | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 8. Related Work
arXiv: 2103.10169
## 8. Related Work
A lot of streaming systems have made their debut in the previous years driving important advances. Some systems, such as Twitter's Storm (Toshniwal et al., 2014) and Heron (Kulkarni et ... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:20"
] | 346 | en | 0.525 | 2026-09-21T09:48:21.602016Z |
arxiv-2103.10169-21 | arxiv | https://arxiv.org/abs/2103.10169 | CC-BY-4.0 | Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 9. Event-Driven Applications as Stateful Dataflow Graphs | Title: Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile — 9. Event-Driven Applications as Stateful Dataflow Graphs
arXiv: 2103.10169
## 9. Event-Driven Applications as Stateful Dataflow Graphs
Jet aims at becoming an execution platform for event-driven applications (Katsifodimos and Fragkoulis, ... | papers | [
"paper",
"arxiv_id:2103.10169",
"relevance:4",
"chunk:21"
] | 403 | en | 0.4 | 2026-09-21T09:48:21.602019Z |
arxiv-2204.04898-01 | arxiv | https://arxiv.org/abs/2204.04898 | CC-BY-4.0 | PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 1 Introduction | Title: PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 1 Introduction
arXiv: 2204.04898
## 1 Introduction
Process mining is a branch of data science that aims to analyze the execution of business processes starting from the event data contained in the information
systems supporting the proc... | papers | [
"paper",
"arxiv_id:2204.04898",
"relevance:4",
"chunk:1"
] | 408 | en | 0.425 | 2026-09-21T09:48:33.841843Z |
arxiv-2204.04898-02 | arxiv | https://arxiv.org/abs/2204.04898 | CC-BY-4.0 | PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 2 Preliminaries | Title: PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 2 Preliminaries
arXiv: 2204.04898
## 2 Preliminaries
This section will first present the NVIDIA RAPIDS framework for GPU-enabled data processing and mining. Then, an overview of the most widely used file formats and data structures for ... | papers | [
"paper",
"arxiv_id:2204.04898",
"relevance:4",
"chunk:2"
] | 613 | en | 0.525 | 2026-09-21T09:48:33.841847Z |
arxiv-2204.04898-03 | arxiv | https://arxiv.org/abs/2204.04898 | CC-BY-4.0 | PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 3 Implementation and Tool | Title: PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 3 Implementation and Tool
arXiv: 2204.04898
## 3 Implementation and Tool
In PM4Py-GPU, we assume an event log to be ingested from a Parquet/CSV file into a CuDF dataframe using the methods available in CuDF. On top of such dataframe, di... | papers | [
"paper",
"arxiv_id:2204.04898",
"relevance:4",
"chunk:3"
] | 903 | en | 0.475 | 2026-09-21T09:48:33.841853Z |
arxiv-2204.04898-04 | arxiv | https://arxiv.org/abs/2204.04898 | CC-BY-4.0 | PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 4 Assessment | Title: PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 4 Assessment
arXiv: 2204.04898
## 4 Assessment
In this section, we want to compare PM4Py-GPU against other libraries/solutions for process mining to evaluate mainstream operations' execution time against significant amounts of data. The... | papers | [
"paper",
"arxiv_id:2204.04898",
"relevance:4",
"chunk:4"
] | 533 | en | 0.4 | 2026-09-21T09:48:33.841856Z |
arxiv-2204.04898-05 | arxiv | https://arxiv.org/abs/2204.04898 | CC-BY-4.0 | PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 5 Related Work | Title: PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining — 5 Related Work
arXiv: 2204.04898
## 5 Related Work
Process Mining on Big Data Architectures:
an integration between process mining techniques and Apache Hadoop has been proposed in [3]. Apache Hadoop does not work in-memory and requires... | papers | [
"paper",
"arxiv_id:2204.04898",
"relevance:4",
"chunk:5"
] | 298 | en | 0.45 | 2026-09-21T09:48:33.841860Z |
arxiv-2209.10785-00 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning | Title: Deep Lake: a Lakehouse for Deep Learning
arXiv: 2209.10785 | License: CC-BY-4.0
Abstract: Traditional data lakes provide critical data infrastructure for analytical workloads by enabling time travel, running SQL queries, ingesting data with ACID transactions, and visualizing petabyte-scale datasets on cloud sto... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:0"
] | 1,155 | en | 0.8 | 2026-09-21T09:48:08.120563Z |
arxiv-2209.10785-01 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 2. Current Challenges | Title: Deep Lake: a Lakehouse for Deep Learning — 2. Current Challenges
arXiv: 2209.10785
## 2. Current Challenges
In this section, we discuss the current and historical challenges of unstructured or complex data management.
### 2.1. Complex Data Types in a Databases
It is generally not recommended to store binary ... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:1"
] | 584 | en | 0.8 | 2026-09-21T09:48:08.120579Z |
arxiv-2209.10785-02 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 3. Tensor Storage Format | Title: Deep Lake: a Lakehouse for Deep Learning — 3. Tensor Storage Format
arXiv: 2209.10785
## 3. Tensor Storage Format
Deep Lake datasets follow columnar storage architecture, with tensors as columns, as shown in Fig. 3. Each tensor is a collection of *chunks* - binary blobs that contain the data samples. An index ... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:2"
] | 1,024 | en | 0.575 | 2026-09-21T09:48:08.120584Z |
arxiv-2209.10785-03 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.1. Ingestion | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.1. Ingestion
arXiv: 2209.10785
### 4.1. Ingestion
#### 4.1.1. Extract
Sometimes metadata might already reside in a relational database. We additionally built an ETL destination connector using Airbyte3 air (2020). The framework allows... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:3"
] | 235 | en | 0.525 | 2026-09-21T09:48:08.120587Z |
arxiv-2209.10785-04 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.2. Version Control | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.2. Version Control
arXiv: 2209.10785
### 4.2. Version Control
Deep Lake also addresses the need for the reproducibility of experiments and compliance with a complete data lineage. Different versions of the dataset exist in the same sto... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:4"
] | 333 | en | 0.45 | 2026-09-21T09:48:08.120591Z |
arxiv-2209.10785-05 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.3. Visualization of Tensors | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.3. Visualization of Tensors
arXiv: 2209.10785
### 4.3. Visualization of Tensors
Data visualization is a crucial part of ML workflows, especially when the data is hard to parse analytically. Fast and seamless visualization allows faster... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:5"
] | 196 | en | 0.35 | 2026-09-21T09:48:08.120594Z |
arxiv-2209.10785-06 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.4. Tensor Query Language | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.4. Tensor Query Language
arXiv: 2209.10785
### 4.4. Tensor Query Language
Querying and balancing datasets is a common step in training deep learning workflows. Typically, this is achieved inside a dataloader using sampling strategies o... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:6"
] | 400 | en | 0.5 | 2026-09-21T09:48:08.120598Z |
arxiv-2209.10785-07 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.5. Materialization | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.5. Materialization
arXiv: 2209.10785
### 4.5. Materialization
Most of the raw data used for deep learning is stored as raw files (compressed in formats like JPEG), either locally or on the cloud. A common way to construct datasets is t... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:7"
] | 272 | en | 0.475 | 2026-09-21T09:48:08.120601Z |
arxiv-2209.10785-08 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.6. Streaming Dataloader | Title: Deep Lake: a Lakehouse for Deep Learning — 4. Deep Lake System Overview › 4.6. Streaming Dataloader
arXiv: 2209.10785
### 4.6. Streaming Dataloader
As datasets become larger, storing and transferring over the network from a remotely distributed storage becomes inevitable. Data streaming enables training models... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:8"
] | 247 | en | 0.45 | 2026-09-21T09:48:08.120604Z |
arxiv-2209.10785-09 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 5. Machine Learning Use Cases | Title: Deep Lake: a Lakehouse for Deep Learning — 5. Machine Learning Use Cases
arXiv: 2209.10785
## 5. Machine Learning Use Cases
In this section, we review the applications of Deep Lake.
A typical scenario in a Deep Learning application starts with
- (1)
A raw set of files that is collected on an object storage b... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:9"
] | 751 | en | 0.575 | 2026-09-21T09:48:08.120608Z |
arxiv-2209.10785-10 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 6. Performance Benchmarks | Title: Deep Lake: a Lakehouse for Deep Learning — 6. Performance Benchmarks
arXiv: 2209.10785
## 6. Performance Benchmarks
In this section, we experimentally demonstrate Deep Lake's performance at scale from the point of ingestion into the format up to training at scale against other dataloaders and formats. We compa... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:10"
] | 579 | en | 0.475 | 2026-09-21T09:48:08.120611Z |
arxiv-2209.10785-11 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 7. Discussion and Limitations | Title: Deep Lake: a Lakehouse for Deep Learning — 7. Discussion and Limitations
arXiv: 2209.10785
## 7. Discussion and Limitations
Deep Lake's primary use cases include (a) Deep Learning Model Training, (b) Data Lineage and Version Control, (c) Data Querying, and Analytics, (d) Data Inspection and Quality Control. We... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:11"
] | 566 | en | 0.575 | 2026-09-21T09:48:08.120614Z |
arxiv-2209.10785-12 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 8. Related Work | Title: Deep Lake: a Lakehouse for Deep Learning — 8. Related Work
arXiv: 2209.10785
## 8. Related Work
Multiple projects have tried to improve upon or create new formats for storing unstructured datasets including TFRecord extending Protobuf Pro (2014), Petastorm Pet (2018) extending Parquet Vohra (2016), Feather fea... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:12"
] | 335 | en | 0.575 | 2026-09-21T09:48:08.120618Z |
arxiv-2209.10785-13 | arxiv | https://arxiv.org/abs/2209.10785 | CC-BY-4.0 | Deep Lake: a Lakehouse for Deep Learning — 9. Conclusion | Title: Deep Lake: a Lakehouse for Deep Learning — 9. Conclusion
arXiv: 2209.10785
## 9. Conclusion
We presented Deep Lake, the lakehouse for deep learning. Deep Lake is designed to help deep learning workflows run as seamlessly as analytical workflows run on Modern Data Stack. Notably, Deep Lake is built to retain pr... | papers | [
"paper",
"arxiv_id:2209.10785",
"relevance:4",
"chunk:13"
] | 211 | en | 0.5 | 2026-09-21T09:48:08.120621Z |
arxiv-2211.10280-00 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams
arXiv: 2211.10280 | License: CC-BY-4.0
Abstract: Online learning (OL) from data streams is an emerging area of research that encompasses numerous challenges from stream processing, machine learning, and networking. Stream-processing platforms, suc... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:0"
] | 1,426 | en | 0.575 | 2026-09-21T09:49:06.922399Z |
arxiv-2211.10280-01 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 2. Background | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 2. Background
arXiv: 2211.10280
## 2. Background
Considering that the real-time training of ANNs in an OL scenario involves multiple areas of research, we give in the following subsections a short summary of the most important concepts and tech... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:1"
] | 1,104 | en | 0.475 | 2026-09-21T09:49:06.922409Z |
arxiv-2211.10280-02 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 3. TensAIR | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 3. TensAIR
arXiv: 2211.10280
## 3. TensAIR
We now introduce the architecture of TensAIR, the first framework for training and predicting in ANNs models in real-time. TensAIR was designed to work in association with stream-processing engines tha... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:2"
] | 1,324 | en | 0.425 | 2026-09-21T09:49:06.922414Z |
arxiv-2211.10280-03 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion
arXiv: 2211.10280
## 4. Experiments & Discussion
To assess TensAIR, we performed experiments to measure its performance on solving prototypical ML problems such as Word2Vec (word embeddings) and CIFAR-10 (image class... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:3"
] | 526 | en | 0.5 | 2026-09-21T09:49:06.922417Z |
arxiv-2211.10280-04 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.1. Convergence Analysis | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.1. Convergence Analysis
arXiv: 2211.10280
### 4.1. Convergence Analysis
We first explored TensAIR's ability to converge by determining if and how DASGD might degrade the quality of the trained model (Figure 3). W... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:4"
] | 342 | en | 0.45 | 2026-09-21T09:49:06.922420Z |
arxiv-2211.10280-06 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.3. Baseline Comparison | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.3. Baseline Comparison
arXiv: 2211.10280
### 4.3. Baseline Comparison
Apart from TensAIR, it is also possible to train ANNs by using Apache Kafka and Flink as message brokers to generate data streams of varying t... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:6"
] | 478 | en | 0.5 | 2026-09-21T09:49:06.922427Z |
arxiv-2211.10280-07 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.4. Sentiment Analysis of COVID19 | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 4. Experiments & Discussion › 4.4. Sentiment Analysis of COVID19
arXiv: 2211.10280
### 4.4. Sentiment Analysis of COVID19
Here, we exemplify the benefits of training an ANN in real-time from streaming data. To this end, we analyze the impact of... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:7"
] | 741 | en | 0.475 | 2026-09-21T09:49:06.922432Z |
arxiv-2211.10280-08 | arxiv | https://arxiv.org/abs/2211.10280 | CC-BY-4.0 | TensAIR: Real-Time Training of Neural Networks from Data-streams — 5. Conclusions | Title: TensAIR: Real-Time Training of Neural Networks from Data-streams — 5. Conclusions
arXiv: 2211.10280
## 5. Conclusions
OL is an emerging area of research which still has not extensively explored the real-time training of ANNs. In this paper, we introduced TensAIR, a novel system for real-time training of ANNs f... | papers | [
"paper",
"arxiv_id:2211.10280",
"relevance:4",
"chunk:8"
] | 320 | en | 0.5 | 2026-09-21T09:49:06.922435Z |
arxiv-2310.08697-00 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More | Title: The Data Lakehouse: Data Warehousing and More
arXiv: 2310.08697 | License: CC-BY-4.0
Abstract: Relational Database Management Systems designed for Online Analytical Processing (RDBMS-OLAP) have been foundational to democratizing data and enabling analytical use cases such as business intelligence and reporting ... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:0"
] | 1,187 | en | 0.525 | 2026-09-21T09:48:02.503711Z |
arxiv-2310.08697-01 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.1. Technical Components (RDBMS-OLAP Components) | Title: The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.1. Technical Components (RDBMS-OLAP Components)
arXiv: 2310.08697
### 2.1. Technical Components (RDBMS-OLAP Components)
The technical components of an RDBMS-OLAP comprise the fundamental building blocks that underpin the in... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:1"
] | 674 | en | 0.475 | 2026-09-21T09:48:02.503721Z |
arxiv-2310.08697-02 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.2. Technical Capabilities provided by RDBMS-OLAP | Title: The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.2. Technical Capabilities provided by RDBMS-OLAP
arXiv: 2310.08697
### 2.2. Technical Capabilities provided by RDBMS-OLAP
RDBMS-OLAP implements a wide range of capabilities to enable efficient data management and analysis. ... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:2"
] | 1,017 | en | 0.5 | 2026-09-21T09:48:02.503725Z |
arxiv-2310.08697-03 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.3. Technology-Independent Practices (DW Practices) | Title: The Data Lakehouse: Data Warehousing and More — 2. Requirements of Data Warehousing › 2.3. Technology-Independent Practices (DW Practices)
arXiv: 2310.08697
### 2.3. Technology-Independent Practices (DW Practices)
Technology-independent data warehousing practices encompass a range of methodologies that priorit... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:3"
] | 603 | en | 0.7 | 2026-09-21T09:48:02.503728Z |
arxiv-2310.08697-04 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 3. Challenges with Data Warehousing | Title: The Data Lakehouse: Data Warehousing and More — 3. Challenges with Data Warehousing
arXiv: 2310.08697
## 3. Challenges with Data Warehousing
- (1)
Limited to structured data workloads: RDBMS-OLAP was primarily designed to store data in a structured form (schema-on-write), so dashboards and analytical applicati... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:4"
] | 492 | en | 0.6 | 2026-09-21T09:48:02.503732Z |
arxiv-2310.08697-05 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 4. Data Lakehouse | Title: The Data Lakehouse: Data Warehousing and More — 4. Data Lakehouse
arXiv: 2310.08697
## 4. Data Lakehouse
A data lakehouse is a type of data architecture that combines the desirable attributes of an RDBMS-OLAP (data warehousing) and a data lake. It mitigates the challenges seen with both these technologies and ... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:5"
] | 428 | en | 0.625 | 2026-09-21T09:48:02.503735Z |
arxiv-2310.08697-06 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.1. Technical Components of a Data Lakehouse | Title: The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.1. Technical Components of a Data Lakehouse
arXiv: 2310.08697
### 5.1. Technical Components of a Data Lakehouse
A data lakehouse addresses the same technical components that we discussed in the data warehousing section (Sec... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:6"
] | 653 | en | 0.8 | 2026-09-21T09:48:02.503738Z |
arxiv-2310.08697-07 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.2. Technical Capabilities | Title: The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.2. Technical Capabilities
arXiv: 2310.08697
### 5.2. Technical Capabilities
This section reviews the technical capabilities of a system that can support data warehousing established in the "Requirements of Data Warehousing"... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:7"
] | 1,193 | en | 0.775 | 2026-09-21T09:48:02.503741Z |
arxiv-2310.08697-08 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.3. Technology-Independent Practices | Title: The Data Lakehouse: Data Warehousing and More — 5. Requirements of Data Lakehousing › 5.3. Technology-Independent Practices
arXiv: 2310.08697
### 5.3. Technology-Independent Practices
The technology-independent practices listed below highlight how a data lakehouse architecture supports the critical practices u... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:8"
] | 679 | en | 0.8 | 2026-09-21T09:48:02.503744Z |
arxiv-2310.08697-09 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 6. Additional Value That Data Lakehousing Brings | Title: The Data Lakehouse: Data Warehousing and More — 6. Additional Value That Data Lakehousing Brings
arXiv: 2310.08697
## 6. Additional Value That Data Lakehousing Brings
Up to this point, we have discussed the various requirements of a data lakehouse architecture that help to achieve similar capabilities as a dat... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:9"
] | 937 | en | 0.8 | 2026-09-21T09:48:02.503749Z |
arxiv-2310.08697-10 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 7. Example of a Lakehouse Implementation | Title: The Data Lakehouse: Data Warehousing and More — 7. Example of a Lakehouse Implementation
arXiv: 2310.08697
## 7. Example of a Lakehouse Implementation
The modular and open nature of a data lakehouse architecture gives us the flexibility to choose the best-of-breed engines and tools depending on the requirement... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:10"
] | 700 | en | 0.725 | 2026-09-21T09:48:02.503754Z |
arxiv-2310.08697-11 | arxiv | https://arxiv.org/abs/2310.08697 | CC-BY-4.0 | The Data Lakehouse: Data Warehousing and More — 8. Conclusion | Title: The Data Lakehouse: Data Warehousing and More — 8. Conclusion
arXiv: 2310.08697
## 8. Conclusion
In this paper, we presented the various components that make up a traditional data warehousing system with the technological component (RDBMS-OLAP) and the technology-independent practices associated with such a pl... | papers | [
"paper",
"arxiv_id:2310.08697",
"relevance:4",
"chunk:11"
] | 195 | en | 0.45 | 2026-09-21T09:48:02.503757Z |
arxiv-2312.16735-00 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms
arXiv: 2312.16735 | License: CC-BY-4.0
Abstract: Existing serverless data analytics systems rely on external storage services like S3 for data shuffling and communication between cloud functions. While this approach provides the elasticity benefits of s... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:0"
] | 356 | en | 0.525 | 2026-09-21T09:47:26.165449Z |
arxiv-2312.16735-01 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 1. Introduction | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 1. Introduction
arXiv: 2312.16735
## 1. Introduction
SQL | SIMD | External Comm. Medium | Hardware | Client Coordinator | Type | Codebase | |
---|---|---|---|---|---|---|---|
Locus (Pu et al., 2019) | No | No | ElastiCache, S3 | x86_64 | Yes | OLAP ... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:1"
] | 990 | en | 0.6 | 2026-09-21T09:47:26.165460Z |
arxiv-2312.16735-02 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 2. Background | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 2. Background
arXiv: 2312.16735
## 2. Background
Stream processing workloads involve continuous data arrivals from diverse sources, necessitating incremental processing. The processing function is unaware of the data stream's start or end points. Con... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:2"
] | 205 | en | 0.55 | 2026-09-21T09:47:26.165464Z |
arxiv-2312.16735-03 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 3. System Architecture | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 3. System Architecture
arXiv: 2312.16735
## 3. System Architecture
Flock is a cloud-native SQL query engine for event-driven analytics on cloud function services. Figure 1 illustrates the Flock's high-level architectural design. The cloud service pro... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:3"
] | 984 | en | 0.625 | 2026-09-21T09:47:26.165467Z |
arxiv-2312.16735-04 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 4. Function Templates | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 4. Function Templates
arXiv: 2312.16735
## 4. Function Templates
### 4.1. Template Specialization
Legal cloud functions are limited to scripts or compiled programs, leading many systems (Perron et al., 2020) to embed the physical plan into the funct... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:4"
] | 1,386 | en | 0.625 | 2026-09-21T09:47:26.165470Z |
arxiv-2312.16735-05 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.1. One-Way Road | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.1. One-Way Road
arXiv: 2312.16735
### 5.1. One-Way Road
Prior work has proposed solutions for data exchange in serverless environments (Klimovic et al., 2018a; Klimovic et al., 2018b; Pu et al., 2019; Perron et al., 2020... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:5"
] | 491 | en | 0.475 | 2026-09-21T09:47:26.165474Z |
arxiv-2312.16735-06 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.2. Sync and Async | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.2. Sync and Async
arXiv: 2312.16735
### 5.2. Sync and Async
When invoking a function asynchronously, AWS Lambda enqueues the event in a Lambda-owned queue and returns immediately, without exposing Lambda's internal queue... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:6"
] | 322 | en | 0.425 | 2026-09-21T09:47:26.165477Z |
arxiv-2312.16735-07 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.3. "No" Coordinator | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.3. "No" Coordinator
arXiv: 2312.16735
### 5.3. "No" Coordinator
Migrating streaming applications from a traditional serverful deployment to a serverless platform presents unique opportunities. In traditional serverful de... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:7"
] | 420 | en | 0.475 | 2026-09-21T09:47:26.165480Z |
arxiv-2312.16735-08 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.4. Function Name | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.4. Function Name
arXiv: 2312.16735
### 5.4. Function Name
The cloud function name consists of three parts: Function Name: <Query Code>-<Query Stage ID>-<Group Member ID> where the query code is the hash digest of a query... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:8"
] | 131 | en | 0.35 | 2026-09-21T09:47:26.165483Z |
arxiv-2312.16735-09 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.5. Function Group | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 5. Function Communication › 5.5. Function Group
arXiv: 2312.16735
### 5.5. Function Group
The cloud function concurrency is the number of instances or execution environments that serve requests at a given time (lam, [n.d.]b). Flock uses put_function_... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:9"
] | 561 | en | 0.575 | 2026-09-21T09:47:26.165486Z |
arxiv-2312.16735-10 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 6. Dynamic Serving Paradigm | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 6. Dynamic Serving Paradigm
arXiv: 2312.16735
## 6. Dynamic Serving Paradigm
Unlike traditional distributed execution engines, Flock's execution plan is a dynamic directed acyclic graph that evolves over time in the cloud, contrasting with static on-... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:10"
] | 894 | en | 0.675 | 2026-09-21T09:47:26.165489Z |
arxiv-2312.16735-11 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation
arXiv: 2312.16735
## 7. Evaluation
Flock is a cloud-native streaming execution engine developed entirely in Rust, comprising 1̃3,000 lines of code (LoC). The SQL component spans 748 LoC, with query operators sourced from the single-host... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:11"
] | 103 | en | 0.4 | 2026-09-21T09:47:26.165492Z |
arxiv-2312.16735-12 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.1. Benchmark | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.1. Benchmark
arXiv: 2312.16735
### 7.1. Benchmark
We evaluate Flock's performance using two streaming benchmarks: the Yahoo Streaming Benchmark (YSB)(Chintapalli et al., 2016) and the NEXMark Benchmark(Tucker et al., 2008). YSB is a... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:12"
] | 205 | en | 0.475 | 2026-09-21T09:47:26.165495Z |
arxiv-2312.16735-13 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.2. Heterogeneous Hardware | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.2. Heterogeneous Hardware
arXiv: 2312.16735
### 7.2. Heterogeneous Hardware
This experiment leverages AVX2 and ARM Neon intrinsics, relying on Rust SIMD auto-vectorization and handwritten Arrow kernels that explicitly employ SIMD in... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:13"
] | 353 | en | 0.475 | 2026-09-21T09:47:26.165498Z |
arxiv-2312.16735-14 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.3. Performance Cost | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.3. Performance Cost
arXiv: 2312.16735
### 7.3. Performance Cost
We evaluated the performance of Flink and Flock using NEXMark queries 1, 2, 3, 5, 7, 10, 11, and 12 in our experiments. N1, N2, and N3 are elementwise queries that feed... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:14"
] | 562 | en | 0.625 | 2026-09-21T09:47:26.165501Z |
arxiv-2312.16735-15 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.4. Invocation Payload | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.4. Invocation Payload
arXiv: 2312.16735
### 7.4. Invocation Payload
Table 2 presents the difference in latency between payload and S3 communication when the Lambda memory size is set to 128MB, focusing on the communication aspect ra... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:15"
] | 424 | en | 0.425 | 2026-09-21T09:47:26.165504Z |
arxiv-2312.16735-16 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.5. Distributed Aggregation | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 7. Evaluation › 7.5. Distributed Aggregation
arXiv: 2312.16735
### 7.5. Distributed Aggregation
Figure 3 presents the latency and billed duration of NEXMark Query 4 (N4) and YSB under both centralized and distributed execution modes. For NEXMark N4, ... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:16"
] | 373 | en | 0.475 | 2026-09-21T09:47:26.165507Z |
arxiv-2312.16735-18 | arxiv | https://arxiv.org/abs/2312.16735 | CC-BY-4.0 | Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 8. Related Work | Title: Flock: A Low-Cost Streaming Query Engine on FaaS Platforms — 8. Related Work
arXiv: 2312.16735
## 8. Related Work
Major cloud providers introduced serverless workflow services(lam, [n.d.]g; azu, [n.d.]a; goo, [n.d.]), which provide easier design and orchestration for serverless workflow applications. Netherite... | papers | [
"paper",
"arxiv_id:2312.16735",
"relevance:6",
"chunk:18"
] | 340 | en | 0.55 | 2026-09-21T09:47:26.165513Z |
arxiv-2403.04570-00 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks
arXiv: 2403.04570 | License: CC-BY-4.0
Abstract: Distributed stream processing frameworks help building scalable and reliable applications that perform transformations and aggregations on continuous... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:0"
] | 1,184 | en | 0.625 | 2026-09-21T09:45:50.243131Z |
arxiv-2403.04570-01 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 2. Background and Related Work | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 2. Background and Related Work
arXiv: 2403.04570
## 2. Background and Related Work
In the following, we briefly introduce the fundamental concepts of modern stream processing frameworks and discu... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:1"
] | 1,168 | en | 0.6 | 2026-09-21T09:45:50.243141Z |
arxiv-2403.04570-02 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 3. Large-Scale Data Shuffling | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 3. Large-Scale Data Shuffling
arXiv: 2403.04570
## 3. Large-Scale Data Shuffling
We illustrate the use case of large-scale data shuffling with requirements for continuous dashboard queries and re... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:2"
] | 369 | en | 0.475 | 2026-09-21T09:45:50.243145Z |
arxiv-2403.04570-03 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.1. Benchmark Design | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.1. Benchmark Design
arXiv: 2403.04570
### 4.1. Benchmark Design
We base our benchmark design on the ACM SIGSOFT Empirical Standard for benchmarking (Ralph et al.... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:3"
] | 250 | en | 0.475 | 2026-09-21T09:45:50.243149Z |
arxiv-2403.04570-04 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.2. The ShuffleBench Task Sample | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.2. The ShuffleBench Task Sample
arXiv: 2403.04570
### 4.2. The ShuffleBench Task Sample
The ShuffleBench task sample consists of a dataflow architecture to be im... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:4"
] | 674 | en | 0.55 | 2026-09-21T09:45:50.243152Z |
arxiv-2403.04570-05 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.3. Qualities, Metrics and Measurement Methods | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.3. Qualities, Metrics and Measurement Methods
arXiv: 2403.04570
### 4.3. Qualities, Metrics and Measurement Methods
According to requirements from industry that ... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:5"
] | 1,055 | en | 0.55 | 2026-09-21T09:45:50.243157Z |
arxiv-2403.04570-06 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.4. Open-Source Benchmark Availability | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 4. The ShuffleBench Benchmark › 4.4. Open-Source Benchmark Availability
arXiv: 2403.04570
### 4.4. Open-Source Benchmark Availability
We provide ShuffleBench as free and open-source research soft... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:6"
] | 146 | en | 0.35 | 2026-09-21T09:45:50.243161Z |
arxiv-2403.04570-07 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation
arXiv: 2403.04570
## 5. Experimental Evaluation
We employ ShuffleBench to experimentally evaluate the performance of Flink (v. 1.17), Hazelcast (v. 5.3), Kafka Streams ... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:7"
] | 234 | en | 0.55 | 2026-09-21T09:45:50.243164Z |
arxiv-2403.04570-08 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.1. Experimental Setup | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.1. Experimental Setup
arXiv: 2403.04570
### 5.1. Experimental Setup
We conduct our experimental evaluation in a Kubernetes cluster managed by the Elastic Kubernetes... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:8"
] | 227 | en | 0.475 | 2026-09-21T09:45:50.243167Z |
arxiv-2403.04570-09 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.2. Baseline Evaluation of Throughput | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.2. Baseline Evaluation of Throughput
arXiv: 2403.04570
### 5.2. Baseline Evaluation of Throughput
In Section 4.3.1, we discuss how the *ad-hoc throughput* method pr... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:9"
] | 667 | en | 0.55 | 2026-09-21T09:45:50.243170Z |
arxiv-2403.04570-10 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.3. Baseline Evaluation of Latency | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.3. Baseline Evaluation of Latency
arXiv: 2403.04570
### 5.3. Baseline Evaluation of Latency
For our baseline latency evaluations, we generate a constant load of 90 ... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:10"
] | 154 | en | 0.425 | 2026-09-21T09:45:50.243173Z |
arxiv-2403.04570-11 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.4. Evaluation of Deployment Impact | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.4. Evaluation of Deployment Impact
arXiv: 2403.04570
### 5.4. Evaluation of Deployment Impact
In our baseline experiments, we deploy 9 instances of the stream proce... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:11"
] | 149 | en | 0.425 | 2026-09-21T09:45:50.243176Z |
arxiv-2403.04570-12 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.5. Evaluation of Record Size Impact | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.5. Evaluation of Record Size Impact
arXiv: 2403.04570
### 5.5. Evaluation of Record Size Impact
In our baseline experiments, we generate data records of 1024-byte s... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:12"
] | 261 | en | 0.525 | 2026-09-21T09:45:50.243179Z |
arxiv-2403.04570-13 | arxiv | https://arxiv.org/abs/2403.04570 | CC-BY-4.0 | ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.6. Evaluation of Consumer Count Impact | Title: ShuffleBench: A Benchmark for Large-Scale Data Shuffling Operations with Distributed Stream Processing Frameworks — 5. Experimental Evaluation › 5.6. Evaluation of Consumer Count Impact
arXiv: 2403.04570
### 5.6. Evaluation of Consumer Count Impact
ShuffleBench allows for configuring the number of real-time co... | papers | [
"paper",
"arxiv_id:2403.04570",
"relevance:6",
"chunk:13"
] | 159 | en | 0.4 | 2026-09-21T09:45:50.243182Z |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.