Spaces:
Configuration error
Configuration error
Update README.md
#1
by Siraj-Wahaj1 - opened
- .gitattributes +35 -0
- .gitignore +0 -26
- README.md +34 -18
- docs/Papers/Analysis/README.md +0 -26
- docs/Papers/Analysis/cost_optimization.md +0 -17
- docs/Papers/Analysis/deep_container.md +0 -18
- docs/Papers/Analysis/euromicro_2022.md +0 -17
- docs/Papers/Analysis/federated_learning_xai.md +0 -17
- docs/Papers/Analysis/google_sre_book.md +0 -17
- docs/Papers/Analysis/gralaf_rca.md +0 -17
- docs/Papers/Analysis/microrca_agent.md +0 -17
- docs/Papers/Analysis/nights_watch.md +0 -17
- docs/Papers/Analysis/tdad_tracing.md +0 -17
- docs/Papers/research_summaries.md +0 -64
- docs/architecture/why_aionyx.md +0 -76
- docs/business/evaluation.md +0 -39
- docs/business/hf_org_card.md +0 -73
- docs/business/hf_org_setup.md +0 -33
- docs/business/licensing_strategy.md +0 -51
- docs/business/platform_integration.md +0 -56
- docs/business/platform_selection.md +0 -52
- docs/thesis/chapter_mapping.md +0 -25
- docs/thesis/literature_review.md +0 -58
- docs/thesis/proposal.md +0 -53
- infra/terraform/kind/aionyx-cluster-config +0 -19
- infra/terraform/kind/main.tf +0 -56
- infra/terraform/kind/observability.tf +0 -26
- infra/terraform/kind/postgres.tf +0 -36
- infra/terraform/kind/tfplan_obs +0 -0
- platform/orchestration/dagster/__init__.py +0 -0
- platform/orchestration/dagster/assets/__init__.py +0 -0
- platform/orchestration/dagster/assets/ingestion.py +0 -31
- platform/orchestration/dagster/assets/ml_models.py +0 -17
- platform/orchestration/dagster/assets/processing.py +0 -23
- platform/orchestration/dagster/definitions.py +0 -11
.gitattributes
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
| 33 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
.gitignore
DELETED
|
@@ -1,26 +0,0 @@
|
|
| 1 |
-
# Terraform
|
| 2 |
-
.terraform/
|
| 3 |
-
*.tfstate
|
| 4 |
-
*.tfstate.backup
|
| 5 |
-
tfplan
|
| 6 |
-
.terraform.lock.hcl
|
| 7 |
-
|
| 8 |
-
# Python
|
| 9 |
-
__pycache__/
|
| 10 |
-
*.py[cod]
|
| 11 |
-
*$py.class
|
| 12 |
-
.venv/
|
| 13 |
-
venv/
|
| 14 |
-
ENV/
|
| 15 |
-
env/
|
| 16 |
-
|
| 17 |
-
# OS
|
| 18 |
-
.DS_Store
|
| 19 |
-
Thumbs.db
|
| 20 |
-
|
| 21 |
-
# Dagster
|
| 22 |
-
storage/
|
| 23 |
-
*.db
|
| 24 |
-
|
| 25 |
-
# AIONYX Local Assets
|
| 26 |
-
docs/assets/branding/logo.png
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
README.md
CHANGED
|
@@ -1,26 +1,42 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
-
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
|
|
|
|
|
|
| 7 |
|
| 8 |
-
##
|
| 9 |
-
- **
|
| 10 |
-
- **
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
- **Infrastructure**: Kubernetes, Terraform
|
| 12 |
-
- **
|
| 13 |
-
- **
|
| 14 |
-
- **ML**: Scikit-learn, PyTorch (Anomaly Detection, Time-series Forecasting)
|
| 15 |
|
| 16 |
-
|
| 17 |
-
- `platform/`: Core application logic (Dagster, ML pipelines, API).
|
| 18 |
-
- `infra/`: Infrastructure as Code (Terraform, K8s manifests).
|
| 19 |
-
- `docs/`: Thesis documentation, system architecture, and business roadmap.
|
| 20 |
-
- `experiments/`: Jupyter notebooks and datasets for research.
|
| 21 |
|
| 22 |
-
##
|
| 23 |
-
|
|
|
|
|
|
|
| 24 |
|
| 25 |
---
|
| 26 |
-
|
|
|
|
| 1 |
+
---
|
| 2 |
+
# This is the Organization Card metadata
|
| 3 |
+
title: AIONYX
|
| 4 |
+
emoji: 📈
|
| 5 |
+
colorFrom: blue
|
| 6 |
+
colorTo: green
|
| 7 |
+
sdk: static
|
| 8 |
+
pinned: false
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# AIONYX Intelligent AIOps for Kubernetes
|
| 12 |
+
|
| 13 |
+
> **Predictive Reliability and Cost Optimization for Cloud-Native Systems.**
|
| 14 |
|
| 15 |
+
AIONYX is an AI-driven platform designed to transform traditional reactive DevOps into proactive **SRE (Site Reliability Engineering)**. Leveraging data-orchestrated state-of-the-art machine learning, AIONYX predicts infrastructure failures and identifies cost-saving opportunities before they impact the business.
|
| 16 |
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
+
## Our Mission
|
| 20 |
+
To bridge the gap between academic Machine Learning research and practical DevOps engineering. AIONYX is built on the philosophy of **"Self-Prevention"**—detecting "Grey Failures" and anomalies using temporal coherence and causal inference.
|
| 21 |
|
| 22 |
+
## Core Research Areas
|
| 23 |
+
- **Real-time Anomaly Detection**: Deep learning for container-level telemetry.
|
| 24 |
+
- **Predictive Resource Scaling**: AI-driven cost optimization for K8s clusters.
|
| 25 |
+
- **Root Cause Analysis (RCA)**: Graph-based liability analysis for microservices.
|
| 26 |
+
- **Explainable AI (XAI)**: Building human trust in autonomous operations.
|
| 27 |
+
|
| 28 |
+
## Our Tech Stack
|
| 29 |
+
- **Orchestration**: Dagster (Data Pipelines)
|
| 30 |
- **Infrastructure**: Kubernetes, Terraform
|
| 31 |
+
- **Database**: PostgreSQL (Feature Store)
|
| 32 |
+
- **AI/ML**: Scikit-Learn, PyTorch, LLM Agents
|
|
|
|
| 33 |
|
| 34 |
+
---
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
|
| 36 |
+
## Links
|
| 37 |
+
- 🐙 **Code & Infrastructure**: [GitHub (aionyx-hq)](https://github.com/aionyx-ops)
|
| 38 |
+
- 📝 **Technical Blog**: *Coming Soon*
|
| 39 |
+
- 🎓 **Thesis Project**: Malmö University (Data Science)
|
| 40 |
|
| 41 |
---
|
| 42 |
+
*AIONYX is currently in active development. We are open to academic collaborations and industry partnerships.*
|
docs/Papers/Analysis/README.md
DELETED
|
@@ -1,26 +0,0 @@
|
|
| 1 |
-
# AIONYX Research Library: Detailed Analysis Index
|
| 2 |
-
|
| 3 |
-
Welcome to the deep-dive section of the AIONYX knowledge base. Here, we analyze each major research paper and book to extract the "Gold" for your Master's Thesis and your business.
|
| 4 |
-
|
| 5 |
-
## 📁 Library Contents
|
| 6 |
-
|
| 7 |
-
### 🛡️ Reliability & SRE Foundations
|
| 8 |
-
- [**Google SRE Book**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/google_sre_book.md): The industry manual for modern cloud operations.
|
| 9 |
-
- [**The Night's Watch Algorithm**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/nights_watch.md): Avoiding alert fatigue with temporal AI.
|
| 10 |
-
|
| 11 |
-
### 🧠 Anomaly Detection & AI
|
| 12 |
-
- [**DeepContainer**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/deep_container.md): Real-time deep learning for K8s containers.
|
| 13 |
-
- [**Federated Learning & XAI**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/federated_learning_xai.md): Building trust and privacy in cloud AI.
|
| 14 |
-
|
| 15 |
-
### 🔍 Root Cause Analysis (RCA)
|
| 16 |
-
- [**MicroRCA-Agent (2025)**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/microrca_agent.md): Using LLM Agents to find and fix bugs.
|
| 17 |
-
- [**GRALAF (GRaph-based Liability Analysis)**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/gralaf_rca.md): Solving the "finger-pointing" problem in microservices.
|
| 18 |
-
|
| 19 |
-
### 💸 Cost Optimization & Business
|
| 20 |
-
- [**Kubernetes Cost Optimization**](file:///Users/sirajulhaqwahaj/my-dir/aionyx/docs/Papers/Analysis/cost_optimization.md): How to save 30-50% on cloud bills using AI.
|
| 21 |
-
|
| 22 |
-
---
|
| 23 |
-
## 🎓 How to use this library for your Thesis
|
| 24 |
-
1. **Literature Review Chapter:** Copy the "Simple Summary" and "Core Technical Concepts" sections to build your citations.
|
| 25 |
-
2. **Implementation Chapter:** Use the "Relevance to AIONYX" sections to justify your technical choices.
|
| 26 |
-
3. **Future Work Chapter:** Look at the "MicroRCA-Agent" and "Federated Learning" docs for inspiration on where to take the project next.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/cost_optimization.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: Kubernetes Cost Optimization
|
| 2 |
-
## *Strategies and Best Practices (Kubegrade Summary)*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
Cloud bills are the #1 nightmare for CFOs. This document explains the "Three Pillars" of saving money on Kubernetes: Right-Sizing, Autoscaling, and Spot Instances.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Right-Sizing:** Most pods are assigned 2GB of RAM but only use 200MB. Reducing this "Requested" resource is the fastest way to save money.
|
| 9 |
-
- **Autoscaling (HPA, VPA, Cluster):** Using the three types of K8s autoscalers together ensure you only pay for what you use *at that exact second*.
|
| 10 |
-
- **FinOps:** Bringing Finance and Engineering teams together to look at the same data.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The "Business Goal":** One of AIONYX's core modules will be a "Cost Predictor." We will use these strategies to give users recommendations like: *"Your API pod is over-provisioned; you could save $200/month by reducing it."*
|
| 14 |
-
- **The "Winning Metric":** In your thesis, don't just measure "Accuracy." Measure **"Dollars Saved."** That is what gets you hired and gets you investors.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
Use this for your "Business Context" or "Cost Evaluation" section. It gives you the "Industry Standard" benchmark (70-80% utilization) to compare your AIONYX results against.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/deep_container.md
DELETED
|
@@ -1,18 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: DeepContainer
|
| 2 |
-
## *Real-time Anomaly Detection in Cloud-Native Container Environments*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
DeepContainer is a "brain" for Kubernetes that watches every container in real-time. It uses Deep Learning (a type of advanced AI) to listen to the "heartbeat" of your containers (CPU, memory, network). If it sees anything that doesn't look like normal behavior, it flags it as an anomaly in milliseconds.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Multi-Layered Detection:** Instead of just looking at one thing, it looks at the container, the pod, and the node all at once.
|
| 9 |
-
- **Low Latency:** It responds in about **7.3ms**, which is faster than a human can blink. This is crucial for stopping attacks or crashes before they spread.
|
| 10 |
-
- **Linear Scalability:** It can handle up to **10,000 containers** without slowing down.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The Blueprint:** This paper provides the mathematical proof that what we are doing with AIONYX is possible at scale.
|
| 14 |
-
- **Benchmarking:** We can use their performance numbers (96.8% accuracy) as a goal for our own AI models.
|
| 15 |
-
- **Feature Selection:** It confirms that tracking resource utilization and network patterns together is the best way to catch "Grey Failures."
|
| 16 |
-
|
| 17 |
-
### 4. Key Takeaway for Your Thesis
|
| 18 |
-
When you write your "Implementation" chapter, you can cite DeepContainer to justify why AIONYX uses real-time stream processing instead of just checking a database every 5 minutes.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/euromicro_2022.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: Euromicro SEAA 2022
|
| 2 |
-
## *Anomaly Detection in Cloud-Native Systems*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
This paper specifically tests **8 different AI algorithms** to see which one is best at catching problems in cloud systems (like Kafka and Zookeeper). It found that AI can be incredibly accurate (over 95%) and very fast to train.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **KPI Dependency:** It studied how 5 major "Key Performance Indicators" depend on 168 smaller metrics.
|
| 9 |
-
- **Algorithm Performance:** It tested 8 models and found that training took only 8-17 minutes—proving that you don't need a supercomputer to run AIOps.
|
| 10 |
-
- **Accuracy:** Accomplished a 95% AUC (Area Under Curve), which is a very high score for real-world production data.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **Metric Selection:** The paper lists **168 metrics**. We can use this list to decide exactly which Postgres columns and Dagster assets to prioritize.
|
| 14 |
-
- **Algorithm Choice:** Since they found that training is fast, it justifies why AIONYX can run its ML models directly on a small Kubernetes node rather than needing a massive expensive server.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
Cite this in your "Data Science" chapter to justify your choice of metrics. It provides the "Scientific Approval" for using standard infrastructure metrics (CPU, Memory, Latency) as the primary inputs for your AI.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/federated_learning_xai.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: Cloud Network Anomaly Detection (2024)
|
| 2 |
-
## *Federated Learning and Explainable AI (XAI)*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
This paper tackles two massive problems: **Privacy** and **Trust**. It uses "Federated Learning," which allows different cloud nodes to learn from each other without ever sharing their actual data. It also uses "Explainable AI" to prove *why* it thinks something is an anomaly.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Federated Learning (FL):** Instead of sending all your logs to one big server, the "brain" (the model) travels to the data. This is great for GDPR and privacy.
|
| 9 |
-
- **Explainable AI (XAI):** It provides a "justification" for its decisions. It doesn't just say "DDoS Attack"; it says "DDoS because Network Packet rate increased by 500%."
|
| 10 |
-
- **Malicious Pattern Recognition:** Specifically designed to catch security threats like ransomware and DDoS in cloud environments.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The "Business Trust" Pillar:** You can use the "Explainable AI" concept to build a dashboard that actually *talks* to the user. *"AIONYX detected an anomaly because CPU spiked simultaneously with a database lock."*
|
| 14 |
-
- **Privacy First:** If we ever sell AIONYX to banks or healthcare, we can mention Federated Learning as a way to protect their sensitive data.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
In your "System Design" chapter, mention that while AIONYX starts centralized, its architecture is built to support Federated Learning in the future for enhanced privacy.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/google_sre_book.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Book Analysis: Site Reliability Engineering (Google)
|
| 2 |
-
## *How Google Runs Production Systems*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
This isn't just a book; it's the "Manual" for modern DevOps. Google invented the term **SRE (Site Reliability Engineering)** because they realized that standard IT operations couldn't scale to billions of users. The core message is: **Treat operations as a software problem.**
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **The Error Budget:** You don't aim for 100% uptime. You decide that 0.1% downtime is okay (the "budget"). You use that time to take risks and deploy new features.
|
| 9 |
-
- **Toil Reduction:** If a human has to do the same task twice, write a script to do it. SREs should spend 50% of their time coding, not just "fixing things."
|
| 10 |
-
- **Monitoring vs. Observability:** You shouldn't just watch "if it's up"; you should watch the "Four Golden Signals": Latency, Traffic, Errors, and Saturation.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The Market Standard:** When AIONYX detects an anomaly, it should report it using the "Four Golden Signals" language. This makes our tool instantly recognizable to any professional SRE.
|
| 14 |
-
- **Business Value:** AIONYX is a tool that **automates toil**. We are selling "time" back to the engineering team so they can focus on their business, not their servers.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
Every term you use in your thesis (SLO, SLI, Error Budget) should be cited from this book. It gives your project "industry legitimacy."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/gralaf_rca.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: GRALAF (ICC 2023)
|
| 2 |
-
## *Graph-Based Liability Analysis for Microservices*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
When a microservice fails, everyone points fingers at each other. This paper introduces "GRALAF," a framework that uses math (Bayesian Networks) to find out who is actually "liable" (responsible) for the failure.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Causal Bayesian Networks (CBN):** It builds a map of "Cause and Effect." If Service A calls Service B, and B is slow, then B is liable for A's slowness.
|
| 9 |
-
- **Fault Injection:** They actually *purposefully* broke the system during testing to see if the AI could find the root cause.
|
| 10 |
-
- **Metric Tracking:** It focuses on high-level metrics like response time and CPU to stay "near-real-time."
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The "Finger-Pointing" Solver:** This is a huge business value. Instead of hours of meetings to find a bug, AIONYX can use this graph logic to say, *"It's not the API's fault; the Database is slow."*
|
| 14 |
-
- **Experiment Methodology:** For your thesis experiments, you can follow their "Fault Injection" method: purposefully stress-test your K8s cluster and see if AIONYX catches it.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
Cite this paper when discussing "Root Cause Analysis." It provides a solid mathematical foundation for why a "Graph-based" approach is better than a simple list of events.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/microrca_agent.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: MicroRCA-Agent (2025)
|
| 2 |
-
## *Microservice Root Cause Analysis using LLM Agents*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
This is the most "cutting-edge" paper in your folder. It asks: *"What if we use an AI like ChatGPT to act as a Senior SRE?"* It uses "Agents" that can actually *reason* about why a system is failing, rather than just pointing at a red graph.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **LLM Reasoning:** The agent reads the logs and says, *"I see a timeout in Service A and a slow query in the Database. Therefore, the Database is the root cause."*
|
| 9 |
-
- **Multimodal Fusion:** It looks at data in different formats (text logs, numeric metrics, and dependency maps) and "fuses" them into one explanation.
|
| 10 |
-
- **Structured Analysis:** It doesn't just give a vague answer; it gives a step-by-step reasoning trace.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The North Star:** This is where we want AIONYX to be in 1-2 years. Your thesis will build the "Data Foundation," and this paper shows how we can add an "AI Agent" on top later.
|
| 14 |
-
- **Human-Centric AIOps:** It proves that business users (Founders/Managers) want explanations in plain English, not just raw data.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
In your "Future Work" or "Conclusion" chapter, you can cite this paper to show that the AIONYX architecture is "Agent-Ready." This shows you are thinking 3-5 years ahead of the current market.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/nights_watch.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: The Night's Watch Algorithm
|
| 2 |
-
## *AI-Driven Anomaly Detection in Cloud-Native Microservices*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
Named after the watchers in "Game of Thrones," this algorithm is designed to stay awake and watch your microservices 24/7. Its main job is to solve "Alert Fatigue"—the problem where engineers get so many fake alerts that they start ignoring the real ones.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Unsupervised Learning:** It doesn't need a human to tell it "this is a crash." It learns what "Wednesday afternoon traffic" looks like vs "Saturday night traffic" automatically.
|
| 9 |
-
- **Temporal Coherence:** It looks at the *order* of events. If A normally happens before B, and suddenly B happens before A, it knows something is wrong.
|
| 10 |
-
- **Multi-Source Data:** It combines logs (text) + metrics (numbers) + alarms to get a full picture.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **The "Truth" Filter:** We will use the Night's Watch philosophy to ensure AIONYX doesn't "cry wolf." This is our #1 selling point for the business: **Trusted Alerts.**
|
| 14 |
-
- **Feature Engineering:** The way they combine logs and metrics is exactly how we should design our Dagster processing assets.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
This paper is perfect for your "Methodology" chapter. You can explain how AIONYX uses temporal coherence to detect subtle anomalies that simple thresholds (like "CPU > 80%") would miss.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/Analysis/tdad_tracing.md
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# Paper Analysis: TDAD (Trace-Driven Anomaly Detection)
|
| 2 |
-
## *Real-Time Anomaly Detection Using Distributed Tracing*
|
| 3 |
-
|
| 4 |
-
### 1. Simple Summary
|
| 5 |
-
Microservices are like a spiderweb. When you click a button, a "Trace" follows the request through 10 different services. This paper uses these traces to find out where the request got stuck or failed.
|
| 6 |
-
|
| 7 |
-
### 2. Core Technical Concepts
|
| 8 |
-
- **Span Causal Graph (SCG):** It turns a technical "Trace" into a visual "Graph." Nodes represent services, and lines represent the calls between them.
|
| 9 |
-
- **Graph Neural Networks (GNN):** It uses a special kind of AI that is very good at understanding shapes and connections (graphs) instead of just lists of numbers.
|
| 10 |
-
- **Root Cause Analysis:** Because it watches the whole request path, it can say exactly which "Span" (the piece of the request) was the slow one.
|
| 11 |
-
|
| 12 |
-
### 3. Relevance to AIONYX
|
| 13 |
-
- **Deep Observability:** While AIONYX starts with "Metrics" (numbers), this paper provides the logic for adding "Tracing" in Phase 2. This is what moves you from a "Junior AIOps" tool to a "Senior Enterprise" tool.
|
| 14 |
-
- **Graph-Based Visualization:** In your dashboard, you can use these "Span Graphs" to show users how an anomaly in the Database slowed down the User Profile service.
|
| 15 |
-
|
| 16 |
-
### 4. Key Takeaway for Your Thesis
|
| 17 |
-
Cite this when explaining why "Metrics" alone aren't enough for microservices. You need to understand the *relationships* between services, not just their CPU usage.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/Papers/research_summaries.md
DELETED
|
@@ -1,64 +0,0 @@
|
|
| 1 |
-
# Research Paper Summaries: The "AIONYX" Knowledge Base
|
| 2 |
-
|
| 3 |
-
I have analyzed the research papers and books found in your `docs/Papers` folder. Below is a simplified explanation of the most important ones and how they help you built **AIONYX**.
|
| 4 |
-
|
| 5 |
-
---
|
| 6 |
-
|
| 7 |
-
## 🔝 The "Must-Reads" for Your Project
|
| 8 |
-
|
| 9 |
-
### 1. Site Reliability Engineering (The Google SRE Book)
|
| 10 |
-
**What it is:** The "Bible" of how Google keeps its systems running.
|
| 11 |
-
**Simple Explanation:** It teaches you that "100% uptime" is a mistake. Instead, you should have an "Error Budget" and use automation (SRE) to manage reliability.
|
| 12 |
-
**Relevance to AIONYX:** This gives you the *language* to talk to customers. You don't just sell "uptime"; you sell "Error Budget Management" and "Toil Reduction."
|
| 13 |
-
|
| 14 |
-
### 2. DeepContainer: Real-time Anomaly Detection
|
| 15 |
-
**What it is:** A paper about using Deep Learning to watch containers in Kubernetes.
|
| 16 |
-
**Simple Explanation:** It uses a "multi-layered" approach to catch security problems and resource exhaustion. It’s very fast (under 10ms) and very accurate.
|
| 17 |
-
**Relevance to AIONYX:** This is your primary competitor/inspiration. It proves that real-time detection in K8s is possible and highly effective.
|
| 18 |
-
|
| 19 |
-
### 3. MicroRCA: Root Cause Analysis
|
| 20 |
-
**What it is:** A famous paper on finding out *why* a microservice failed.
|
| 21 |
-
**Simple Explanation:** It looks at the dependencies between services (like a map) to find the "patient zero" of a failure.
|
| 22 |
-
**Relevance to AIONYX:** You can use this logic in your "Root Cause" dashboard to tell users exactly which pod started the problem.
|
| 23 |
-
|
| 24 |
-
### 4. The Night's Watch Algorithm (applsci-15-12762)
|
| 25 |
-
**What it is:** A very recent (2025 style) AI-driven algorithm for microservices.
|
| 26 |
-
**Simple Explanation:** It's called "unsupervised," meaning it learns what "normal" looks like on its own without you having to tell it. It reduces "False Alarms" (annoying alerts that aren't real problems).
|
| 27 |
-
**Relevance to AIONYX:** This is a key feature for your business. Customers hate annoying alerts. The "Night's Watch" approach helps AIONYX be a "quiet" and "trusted" advisor.
|
| 28 |
-
|
| 29 |
-
### 5. MicroRCA-Agent (2509.15635v1 - The Latest Tech)
|
| 30 |
-
**What it is:** Using **Large Language Models (LLMs)** like me to do the analysis.
|
| 31 |
-
**Simple Explanation:** It uses "Agents" to read logs and metrics together and then writes a report in plain English explaining the fault.
|
| 32 |
-
**Relevance to AIONYX:** This is the future of your project! Instead of just a graph, AIONYX can provide a "Human-like" report to the SRE: *"Hey, I think Pod-B is failing because the database is slow."*
|
| 33 |
-
|
| 34 |
-
---
|
| 35 |
-
|
| 36 |
-
## 💰 The Business Value Papers
|
| 37 |
-
|
| 38 |
-
### Kubernetes Cost Optimization (Kubegrade Summary)
|
| 39 |
-
**Simple Explanation:** Focuses on "Right-Sizing." Most companies buy "Extra Large" servers but only use 10% of them.
|
| 40 |
-
**Relevance to AIONYX:** This is how you make money. Your thesis will show that AIONYX "recommends" the perfect size, saving the company 30-50% on their bill.
|
| 41 |
-
|
| 42 |
-
---
|
| 43 |
-
|
| 44 |
-
## 🛠 What to look for when you study these:
|
| 45 |
-
|
| 46 |
-
1. **"Grey Failures":** Look for how they find problems where the pod is still "Running" but the user is unhappy (slow latency).
|
| 47 |
-
2. **"Causal Graphs":** Look at how they draw lines between services to show who depends on who.
|
| 48 |
-
3. **"Precision vs. Recall":** In your thesis, don't just say "It works." Use these terms. *Precision* means "When I say it's an anomaly, am I right?" *Recall* means "Did I catch ALL the anomalies?"
|
| 49 |
-
|
| 50 |
-
---
|
| 51 |
-
|
| 52 |
-
### 📝 Summary of Relevance Table
|
| 53 |
-
|
| 54 |
-
| Topic | Paper/Book | Use for AIONYX |
|
| 55 |
-
| :--- | :--- | :--- |
|
| 56 |
-
| **Terminology** | Google SRE Book | Writing your Introduction. |
|
| 57 |
-
| **Detection Logic** | DeepContainer / Night's Watch | Defining your ML models. |
|
| 58 |
-
| **Root Cause** | MicroRCA / GRALAF | Designing your Dashboard. |
|
| 59 |
-
| **The "Future"** | MicroRCA-Agent (LLMs) | Making AIONYX a premium product. |
|
| 60 |
-
| **Money/Value** | Kubegrade | Writing your Business Case. |
|
| 61 |
-
|
| 62 |
-
---
|
| 63 |
-
|
| 64 |
-
**Professor's Tip:** You don't need to read every page of these 17 PDFs. Read the **Abstract**, the **Introduction**, and the **Conclusion** first. If it feels relevant, look at the **Methodology** section to see what metrics (CPU, RAM, Latency) they used.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/architecture/why_aionyx.md
DELETED
|
@@ -1,76 +0,0 @@
|
|
| 1 |
-
# Why AIONYX? Reactive Self-Healing vs. Predictive Intelligence
|
| 2 |
-
|
| 3 |
-
A common question from Kubernetes engineers is: *"If Kubernetes ensures my desired state is running and cloud providers guarantee control-plane availability, why do I need AIONYX?"*
|
| 4 |
-
|
| 5 |
-
This document addresses the critical gap between **Status Quo (K8s)** and **Advanced Ops (AIONYX)**.
|
| 6 |
-
|
| 7 |
-
## 1. The "Reconciliation Loop" Fallacy
|
| 8 |
-
Kubernetes is famous for its self-healing: if a pod crashes, it restarts it. However, **restarts are reactive.**
|
| 9 |
-
|
| 10 |
-
- **K8s (Reactive):** Pod dies → 2 minutes of downtime → K8s detects failure → K8s restarts pod. Customer experiences an error.
|
| 11 |
-
- **AIONYX (Proactive):** Predicts memory leak → 15 minutes before the crash → AIONYX triggers a graceful canary rollout or scales the node. Customer experiences **zero** errors.
|
| 12 |
-
|
| 13 |
-
> [!IMPORTANT]
|
| 14 |
-
> AIONYX transforms "Self-Healing" into "Self-Prevention."
|
| 15 |
-
|
| 16 |
-
## 2. "Running" vs. "Healthy"
|
| 17 |
-
A Kubernetes Pod can be in a `Running` state while being functionally broken or extremely slow.
|
| 18 |
-
- **K8s:** Sees the process is alive. Liveness probe passes. Everything is "green."
|
| 19 |
-
- **AIONYX:** Uses ML to see that the *latency* of this pod is 3 standard deviations higher than normal for this time of day. It detects a "Grey Failure" that standard K8s would ignore.
|
| 20 |
-
|
| 21 |
-
## 3. The Cost of Managed Availability
|
| 22 |
-
Cloud providers (EKS/GKE) ensure the K8s API is up, but they charge you for every CPU cycle you reserve.
|
| 23 |
-
- **Standard K8s Ops:** To ensure availability, teams "Over-provision" (e.g., reserving 2x more memory than needed) just in case a spike happens.
|
| 24 |
-
- **AIONYX:** Predicts upcoming traffic spikes using historical data. It allows you to run **Leaner Clusters** with lower safety margins because it can scale *before* the spike hits, saving 20-40% on monthly cloud bills.
|
| 25 |
-
|
| 26 |
-
## 4. The "Replica Illusion": Why Replicas Aren't Enough
|
| 27 |
-
A senior engineer might argue: *"I have 3 replicas. If one dies, K8s routes to the others. The user sees nothing."*
|
| 28 |
-
|
| 29 |
-
This is the **"Healthy Replica Illusion."** Here is why AIONYX is still critical:
|
| 30 |
-
|
| 31 |
-
- **The Cascading Failure (Thundering Herd):** If Pod A dies because of a memory spike, the traffic it was handling is immediately pushed to Pod B and C. Now B and C have **50% more load.** If they were already near their limit, they will also crash. This is how a single pod death turns into a full cluster outage. AIONYX predicts the trend and scales *up* before the first pod even hits the limit.
|
| 32 |
-
- **Broken Sessions/Transactions:** Even with 100 replicas, the specific user who was connected to the dying pod will experience a **disrupted connection.** Their checkout might fail, or their file upload might corrupt. AIONYX enables "Graceful Drain"—predicting failure and moving users *before* the pod terminates.
|
| 33 |
-
- **The "Cold Start" Latency:** When K8s restarts a pod, it takes time to "warm up" (JVM startup, cache loading). During this window, your cluster is running at reduced capacity, leading to global latency spikes.
|
| 34 |
-
|
| 35 |
-
## 5. Root Cause vs. Symptom Management
|
| 36 |
-
When a pod crashes repeatedly (CrashLoopBackOff), K8s blindly attempts to restart it. It doesn't tell you *why*.
|
| 37 |
-
- **K8s:** Manages the **symptom** (the crash).
|
| 38 |
-
- **AIONYX:** Analyzes the correlation between log errors and metric spikes across the cluster. It provides the **Root Cause** to the developer via the dashboard, reducing MTTR (Mean Time To Recovery).
|
| 39 |
-
|
| 40 |
-
## 5. Summary Table: The AIONYX Advantage
|
| 41 |
-
|
| 42 |
-
| Feature | Standard K8s + Cloud | AIONYX Platform |
|
| 43 |
-
| :--- | :--- | :--- |
|
| 44 |
-
| **Philosophy** | Reactive (Fix after failure) | Predictive (Avoid failure) |
|
| 45 |
-
| **Availability** | High (Restarts) | Ultra-High (Pre-warnings) |
|
| 46 |
-
| **Cost** | Fixed/Wasteful (Over-provisioning) | Optimized (ML-driven scaling) |
|
| 47 |
-
| **Intelligence** | Rule-based (Triggers/Probes) | Pattern-based (ML Inference) |
|
| 48 |
-
| **Visibility** | Metrics/Logs isolated | Correlated Insights |
|
| 49 |
-
|
| 50 |
-
## 6. Target Customers: Who needs AIONYX?
|
| 51 |
-
|
| 52 |
-
AIONYX is tailor-made for companies where **reliability is non-negotiable** but **SRE headcount is limited.**
|
| 53 |
-
|
| 54 |
-
* **Fintech & Payment Startups:** Where every failed transaction is a compliance or revenue disaster and high uptime is a legal requirement.
|
| 55 |
-
* **E-commerce Platforms:** Teams facing massive seasonal spikes (Black Friday) who currently "over-provision" to survive and waste thousands in the process.
|
| 56 |
-
* **Mid-Scale SaaS (20-100 Nodes):** Companies that have outgrown basic Grafana dashboards but aren't large enough to build their own internal AIOps platforms.
|
| 57 |
-
|
| 58 |
-
## 7. Marketing to Developers and Founders
|
| 59 |
-
|
| 60 |
-
To win this market, we don't sell "AI"—we sell **"Peace of Mind."**
|
| 61 |
-
|
| 62 |
-
1. **The "Cloud Waste" Audit:** Marketing tool that connects to their cluster, runs a Dagster analysis, and shows them exactly how many dollars they are wasting on "Safety Margins" right now.
|
| 63 |
-
2. **The "MTTR" Killer:** Focus on how AIONYX gives you the root cause *before* the SRE is甚至 even paged.
|
| 64 |
-
3. **Open Source "Lead Magnet":** We give away the ingestion code (Dagster assets) for free. This builds the brand. The "AI Brain" (Managed Inference) is what they pay for.
|
| 65 |
-
|
| 66 |
-
## 8. Building Trust in an AI-Driven System
|
| 67 |
-
|
| 68 |
-
Engineers are naturally skeptical of "Magic AI" managing their infra. AIONYX builds trust through **Transparency:**
|
| 69 |
-
|
| 70 |
-
* **Explainable AI (XAI):** When AIONYX predicts a failure, it doesn't just say "Anomaly." It shows a "Contribution Map" (e.g., *70% weight on Memory Increase, 20% on Disk I/O, 10% on Latency*).
|
| 71 |
-
* **Advise-before-Act:** In the first 30 days, AIONYX only *recommends* actions. Once the team sees the predictions are 95% accurate, they flip the switch to "Auto-Pilot."
|
| 72 |
-
* **Academic Foundations:** Transparency about the research. The system isn't a "Black Box"; it's a peer-reviewed methodology from your Master's Thesis.
|
| 73 |
-
|
| 74 |
-
---
|
| 75 |
-
## Conclusion
|
| 76 |
-
Kubernetes is the **Operating System** of the cloud. AIONYX is the **Intelligence** that makes that OS run efficiently, cost-effectively, and proactively. While K8s handles the "Desired State," AIONYX handles the **"Optimal State."**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/evaluation.md
DELETED
|
@@ -1,39 +0,0 @@
|
|
| 1 |
-
# AIONYX: Strategic Evaluation & Business Case
|
| 2 |
-
|
| 3 |
-
As your professor and a technical advisor, I have evaluated **AIONYX** from three critical angles: **Academic Rigor**, **Technical Excellence**, and **Business Viability**.
|
| 4 |
-
|
| 5 |
-
## 1. Academic Perspective (The Thesis Angle)
|
| 6 |
-
**Rating: 4.5/5**
|
| 7 |
-
- **Novelty**: While anomaly detection in K8s isn't brand new, using **software-defined assets (Dagster)** for a reproducible MLOps lifecycle in SRE (Site Reliability Engineering) is a fresh and high-value research direction.
|
| 8 |
-
- **Reproducibility**: This is your strongest point. Most theses fail because they can't prove their data stayed clean. Dagster guarantees this.
|
| 9 |
-
- **Data Science Depth**: You aren't just "running a model"; you are building a **Feature Store** for infrastructure. This moves the project from "DevOps Engineering" to "Data Science Research."
|
| 10 |
-
|
| 11 |
-
## 2. Technical Perspective (The Engineering Angle)
|
| 12 |
-
**Rating: 5/5**
|
| 13 |
-
- **Architecture**: Separating ingestion, processing, and ML into Dagster assets makes the system modular. You can replace the "Anomaly Detector" with a "Cost Predictor" without changing the infrastructure.
|
| 14 |
-
- **Scalability**: By using Postgres (and potentially TimescaleDB) and K8s, your foundation can theoretically handle thousands of nodes.
|
| 15 |
-
- **Tooling Choice**: K8s + Terraform + Dagster is the "Gold Standard" for modern cloud-native AI.
|
| 16 |
-
|
| 17 |
-
## 3. Business Perspective (The Startup Angle)
|
| 18 |
-
**Rating: 4/5**
|
| 19 |
-
### Why it can succeed:
|
| 20 |
-
1. **The "Downtime is Expensive" Problem**: Every minute of downtime costs companies thousands. A tool that *predicts* failure before it happens is an easy sell.
|
| 21 |
-
2. **Cloud Cost Optimization**: Most companies waste 30% of their cloud budget. Predicting when to scale *down* using ML (not just static rules) is a huge value proposition.
|
| 22 |
-
3. **The Data Moat**: As you collect more metrics and labels, your ML models become more accurate than any "generic" tool. This is your competitive advantage.
|
| 23 |
-
|
| 24 |
-
### The Challenges (Risks):
|
| 25 |
-
1. **Data Noise**: Kubernetes labels and metrics can be incredibly noisy. Your feature engineering must be world-class to avoid "alert fatigue."
|
| 26 |
-
2. **Competition**: Datadog and Dynatrace have deep pockets. To win, AIONYX must be **easier to set up** and **better at predicting** specifically for K8s.
|
| 27 |
-
|
| 28 |
-
## 🚀 Can it be a successful business?
|
| 29 |
-
**YES.**
|
| 30 |
-
|
| 31 |
-
If you view AIONYX not as a "monitoring tool" but as an **"Autonomous SRE"**, there is a massive market.
|
| 32 |
-
|
| 33 |
-
**The Winning Strategy**:
|
| 34 |
-
- Start as an **Open Source** tool for the community (your thesis code).
|
| 35 |
-
- Build a **Cloud SaaS** version that manages the Dagster pipelines and Postgres for users.
|
| 36 |
-
- Focus on a specific niche (e.g., "AI-driven cost reduction for Fintech startups on K8s").
|
| 37 |
-
|
| 38 |
-
---
|
| 39 |
-
**Verdict**: You have hit the "Sweet Spot". You are solving a real-world problem with modern tools, and you are doing it in a way that is academically defensible. This is the definition of a high-impact Master's project.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/hf_org_card.md
DELETED
|
@@ -1,73 +0,0 @@
|
|
| 1 |
-
# AIONYX: Hugging Face Organization Card (README)
|
| 2 |
-
|
| 3 |
-
To complete your profile on Hugging Face, you need to create an **Organization Card**. This is the first thing people see when they visit `huggingface.co/aionyx`.
|
| 4 |
-
|
| 5 |
-
## 🛠 How to create it
|
| 6 |
-
1. On your Organization page, click **"New Space"**.
|
| 7 |
-
2. **Space Name:** Must be exactly `README`.
|
| 8 |
-
3. **License:** Select **Apache 2.0**.
|
| 9 |
-
4. **Select the Space SDK:** Choose **Static** (usually under "Choose a Static template").
|
| 10 |
-
5. **Visibility:** Select **Public**.
|
| 11 |
-
6. Click **Create Space**.
|
| 12 |
-
7. Once created, edit the `README.md` file in that space and paste the template below.
|
| 13 |
-
|
| 14 |
-
### Why Static? (vs Gradio or Docker)
|
| 15 |
-
- **Static (Recommended):** This is for pure Markdown/HTML content. It’s perfect for a README or Organization Card. It’s the fastest to load, uses zero "compute nodes" (it's free and always on), and is precisely what a landing page should be.
|
| 16 |
-
- **Gradio / Streamlit:** These are for interactive Python demos (like a chatbot UI). While great for your *future* anomaly detection demo, it’s unnecessary "bloat" for an "About Us" page.
|
| 17 |
-
- **Docker:** This is for full, custom applications. It’s much harder to set up and maintain. You would only use this if you were hosting the entire AIONYX dashboard on Hugging Face.
|
| 18 |
-
|
| 19 |
-
> [!TIP]
|
| 20 |
-
> Use **Static** for the `README` repository. Save **Gradio** for when you want to build a live demo of your ML models later!
|
| 21 |
-
|
| 22 |
-
## 📝 Recommended Content (Copy-Paste)
|
| 23 |
-
|
| 24 |
-
Below is a professional, high-impact template tailored for AIONYX.
|
| 25 |
-
|
| 26 |
-
```markdown
|
| 27 |
-
---
|
| 28 |
-
title: AIONYX
|
| 29 |
-
emoji: 📈
|
| 30 |
-
colorFrom: blue
|
| 31 |
-
colorTo: green
|
| 32 |
-
sdk: static
|
| 33 |
-
pinned: false
|
| 34 |
-
---
|
| 35 |
-
|
| 36 |
-
# 💎 AIONYX: Intelligent AIOps for Kubernetes
|
| 37 |
-
|
| 38 |
-
> **Predictive Reliability and Cost Optimization for Cloud-Native Systems.**
|
| 39 |
-
|
| 40 |
-
AIONYX is an AI-driven platform designed to transform traditional reactive DevOps into proactive **SRE (Site Reliability Engineering)**. Leveraging data-orchestrated state-of-the-art machine learning, AIONYX predicts infrastructure failures and identifies cost-saving opportunities before they impact the business.
|
| 41 |
-
|
| 42 |
-
---
|
| 43 |
-
|
| 44 |
-
## 🧬 Our Mission
|
| 45 |
-
To bridge the gap between academic Machine Learning research and practical DevOps engineering. AIONYX is built on the philosophy of **"Self-Prevention"**—detecting "Grey Failures" and anomalies using temporal coherence and causal inference.
|
| 46 |
-
|
| 47 |
-
## 🧠 Core Research Areas
|
| 48 |
-
- **Real-time Anomaly Detection**: Deep learning for container-level telemetry.
|
| 49 |
-
- **Predictive Resource Scaling**: AI-driven cost optimization for K8s clusters.
|
| 50 |
-
- **Root Cause Analysis (RCA)**: Graph-based liability analysis for microservices.
|
| 51 |
-
- **Explainable AI (XAI)**: Building human trust in autonomous operations.
|
| 52 |
-
|
| 53 |
-
## 🛠 Our Tech Stack
|
| 54 |
-
- **Orchestration**: Dagster (Data Pipelines)
|
| 55 |
-
- **Infrastructure**: Kubernetes, Terraform
|
| 56 |
-
- **Database**: PostgreSQL (Feature Store)
|
| 57 |
-
- **AI/ML**: Scikit-Learn, PyTorch, LLM Agents
|
| 58 |
-
|
| 59 |
-
---
|
| 60 |
-
|
| 61 |
-
## 🔗 Links
|
| 62 |
-
- 🐙 **Code & Infrastructure**: [GitHub (aionyx-hq)](https://github.com/aionyx-ops)
|
| 63 |
-
- 📝 **Technical Blog**: *Coming Soon*
|
| 64 |
-
- 🎓 **Thesis Project**: Malmö University (Data Science)
|
| 65 |
-
|
| 66 |
-
---
|
| 67 |
-
*AIONYX is currently in active development. We are open to academic collaborations and industry partnerships.*
|
| 68 |
-
```
|
| 69 |
-
|
| 70 |
-
## 💡 Why this works
|
| 71 |
-
- **Professionalism**: It uses a "brand quote" at the top to instantly explain the value.
|
| 72 |
-
- **Keywords**: Includes "SRE," "Grey Failures," and "Explainable AI," which attracts the right industry people.
|
| 73 |
-
- **Academic Merit**: Specifically mentions the Master's Thesis at Malmö University, giving the project credibility.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/hf_org_setup.md
DELETED
|
@@ -1,33 +0,0 @@
|
|
| 1 |
-
# AIONYX: Hugging Face Organization Profile Guide
|
| 2 |
-
|
| 3 |
-
To make **AIONYX** look professional to both your thesis reviewers and future investors, follow this guide to fill out your Hugging Face Organization profile.
|
| 4 |
-
|
| 5 |
-
## 📝 Profile Fields Guide
|
| 6 |
-
|
| 7 |
-
| Field | Recommended Value | Why? |
|
| 8 |
-
| :--- | :--- | :--- |
|
| 9 |
-
| **Organization Username** | `aionyx` (or `aionyx-ai`) | Matches your GitHub brand. |
|
| 10 |
-
| **Organization Type** | **Company** | Shows professional intent. |
|
| 11 |
-
| **Homepage** | `https://github.com/aionyx-hq` (for now) | Use your GitHub Org URL until you have a landing page. |
|
| 12 |
-
| **Organization Full name** | `AIONYX: Intelligent AIOps Platform` | Clear, descriptive, and authoritative. |
|
| 13 |
-
| **GitHub username** | `aionyx-hq` | Links your code and models together. |
|
| 14 |
-
| **Twitter / LinkedIn** | *Leave blank for now* | Add these once you launch the business side. |
|
| 15 |
-
| **AI & ML Interests** | `AIOps, Kubernetes Reliability, Time-series Anomaly Detection, Root Cause Analysis, Explainable AI (XAI)` | These keywords help people find your research and product. |
|
| 16 |
-
|
| 17 |
-
---
|
| 18 |
-
|
| 19 |
-
## 🎨 Branding & Logo
|
| 20 |
-
A professional logo is the first step in building a "Brand Moat."
|
| 21 |
-
|
| 22 |
-

|
| 23 |
-
|
| 24 |
-
### Why this design?
|
| 25 |
-
- **Onyx Colors**: Deep blacks and greys represent stability and the "Onyx" part of the name.
|
| 26 |
-
- **Vibrant Accents**: Cyan/Blue represents the "Intelligence" and "Cloud" (Kubernetes) nature of the project.
|
| 27 |
-
- **Minimalism**: High-end startups use clean, recognizable symbols that look good on a dashboard or a dark-mode IDE.
|
| 28 |
-
|
| 29 |
-
---
|
| 30 |
-
|
| 31 |
-
## 🚀 Final Tip for the Profile
|
| 32 |
-
In the **"About"** section (if available), use this single, high-impact sentence:
|
| 33 |
-
> "AIONYX is an AI-driven AIOps platform that leverages data-orchestrated machine learning to predict failures and optimize costs in cloud-native Kubernetes environments."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/licensing_strategy.md
DELETED
|
@@ -1,51 +0,0 @@
|
|
| 1 |
-
# AIONYX: Licensing & Intellectual Property Strategy
|
| 2 |
-
|
| 3 |
-
Choosing the right license is a balance between **Academic Openness** (for your thesis) and **Business Protection** (for your startup).
|
| 4 |
-
|
| 5 |
-
## 1. Recommendation: Apache License 2.0
|
| 6 |
-
I recommend that AIONYX uses the **Apache License 2.0**.
|
| 7 |
-
|
| 8 |
-
### Why Apache 2.0?
|
| 9 |
-
1. **The "Industry Standard":** Kubernetes, Terraform, and Dagster are all either Apache 2.0 or similarly permissive. Using this license makes you look "at home" in the DevOps ecosystem.
|
| 10 |
-
2. **Patent Protection:** Unlike the MIT license, Apache 2.0 specifically includes a "Patent Grant." This prevents people from contributing code to your project and then later suing you for patent infringement.
|
| 11 |
-
3. **Business Friendly:** It allows companies to use your code and even modify it for their internal use without being forced to share their secret modifications (unlike GPL). Small startups and big enterprises trust Apache 2.0.
|
| 12 |
-
|
| 13 |
-
### Why not the others?
|
| 14 |
-
| License | Why Not? |
|
| 15 |
-
| :--- | :--- |
|
| 16 |
-
| **MIT** | It is too simple. It doesn't have the robust language about patents that a business-focused project like AIONYX needs. |
|
| 17 |
-
| **GPL (v3)** | This is a "Copyleft" license. If a company uses your GPL code, they might be legally forced to open-source *their* entire product. This scares away many enterprise customers. |
|
| 18 |
-
| **BSL (Business Source License)** | This is used by companies like Terraform (HashiCorp) to prevent cloud giants from selling their project as a service. It's too complex and "aggressive" for a Master's thesis. |
|
| 19 |
-
|
| 20 |
-
## 2. What if I use "No License"? (All Rights Reserved)
|
| 21 |
-
You might be tempted to leave the project without a license for now. Here is what happens legally and practically:
|
| 22 |
-
|
| 23 |
-
### Theoretical Reality: All Rights Reserved
|
| 24 |
-
- **GitHub & Hugging Face:** If you do not include a license file, you retain full copyright. This means **nobody can legally copy, distribute, or modify your code/models.**
|
| 25 |
-
- **The "Look but don't touch" Rule:** People can read your code on GitHub, but they cannot use it in their own projects without your explicit permission.
|
| 26 |
-
|
| 27 |
-
### Why "No License" is risky for AIONYX:
|
| 28 |
-
1. **Academia (The Thesis):** A core part of a Data Science thesis is **Reproducibility**. If your examiner cannot legally run your code to verify your results, it might raise academic red flags.
|
| 29 |
-
2. **Business Trust:** Professional DevOps teams and investors are afraid of "No License" code. Their legal departments will block them from even testing your tool because the risk of a future copyright lawsuit is too high.
|
| 30 |
-
3. **Hugging Face Discovery:** Hugging Face acts as a search engine for models. Models without licenses are often ranked lower in search results because the platform wants to promote "usable" AI.
|
| 31 |
-
|
| 32 |
-
> [!WARNING]
|
| 33 |
-
> Choosing "No License" is actually more restrictive than a "Business License." It acts as a "Stop Sign" for everyone, including potential customers.
|
| 34 |
-
|
| 35 |
-
## 3. The "Open Core" Business Model
|
| 36 |
-
Since you are a student on a budget, your business strategy should likely be **"Open Core."**
|
| 37 |
-
|
| 38 |
-
1. **The Open Source Core (Apache 2.0):** You release the data ingestion (Dagster assets), the infrastructure (Terraform), and the basic anomaly detection models for free. This builds your reputation.
|
| 39 |
-
2. **The Proprietary "Pro" Layer:** You keep the advanced "Reasoning Agents" (like the MicroRCA-Agent from the papers) and the Managed Dashboard as **Private Code**. Customers pay for this extra layer of intelligence.
|
| 40 |
-
|
| 41 |
-
---
|
| 42 |
-
|
| 43 |
-
## 3. Intellectual Property (IP) Considerations
|
| 44 |
-
As a student at Malmö University, you should double-check the university's policy on IP. However, in most Swedish universities (the "Professor's Privilege"), the student or researcher owns 100% of their IP.
|
| 45 |
-
|
| 46 |
-
- **Action:** Ensure every code file contains a small header:
|
| 47 |
-
`Copyright 2025 [Your Name] / AIONYX. Licensed under Apache 2.0.`
|
| 48 |
-
- **Contributor License Agreement (CLA):** If you ever allow other people to contribute to AIONYX, have them sign a simple CLA. This ensures you keep the right to own the business vision.
|
| 49 |
-
|
| 50 |
-
---
|
| 51 |
-
**Verdict**: Start with **Apache License 2.0**. It is professional, safe for business, and perfectly suited for an academic thesis in the Cloud/DevOps space.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/platform_integration.md
DELETED
|
@@ -1,56 +0,0 @@
|
|
| 1 |
-
# AIONYX: Integration Guide (GitHub & Hugging Face)
|
| 2 |
-
|
| 3 |
-
To build a professional identity for AIONYX, you need to sync your **Code (GitHub)** with your **Models (Hugging Face)**. Here is how to do it efficiently without creating unnecessary accounts.
|
| 4 |
-
|
| 5 |
-
## 1. Do you need a new Hugging Face Account?
|
| 6 |
-
**No.** You do not need a new personal login.
|
| 7 |
-
- **The Strategy:** Much like GitHub, Hugging Face has **Organizations**.
|
| 8 |
-
- **Action:** Keep your personal account, but create a new **Organization** named `aionyx` or `aionyx-ai`.
|
| 9 |
-
- **Benefit:** You can be the "Owner" of the AIONYX organization using your existing account. This keeps your personal profile and your professional work separate but managed from one login.
|
| 10 |
-
|
| 11 |
-
### Which Organization Type to choose?
|
| 12 |
-
When creating the organization on Hugging Face, you will see several options. Here is the best choice for your situation:
|
| 13 |
-
|
| 14 |
-
| Type | Choice | Why? |
|
| 15 |
-
| :--- | :---: | :--- |
|
| 16 |
-
| **Company** | **Recommended** | This is the most professional option. It signals to investors and customers that AIONYX is a serious venture. You don't need a registered legal entity yet to select this. |
|
| 17 |
-
| **Community** | No | Good for general open-source groups, but less "startup-ready." |
|
| 18 |
-
| **University** | No | AIONYX is your personal project, not an official university department. |
|
| 19 |
-
| **Non-Profit** | No | Only if you don't plan to ever make money from it. |
|
| 20 |
-
|
| 21 |
-
> [!TIP]
|
| 22 |
-
> Choose **Company**. It allows you to build a brand now that you can legally register later without changing your Hugging Face identity.
|
| 23 |
-
There are two ways to connect them, depending on your workflow:
|
| 24 |
-
|
| 25 |
-
### A. The "Syncing" Method (Recommended)
|
| 26 |
-
You can set up your Hugging Face space or model repository to automatically "mirror" a GitHub repository.
|
| 27 |
-
1. **Create a Repository** in your GitHub Organization (e.g., `aionyx-hq/anomaly-detection`).
|
| 28 |
-
2. **Create a Model/Space** in your Hugging Face Organization.
|
| 29 |
-
3. **Link them:** In Hugging Face settings, go to "Connected Accounts" or use the "GitHub Sync" feature to automatically pull code from GitHub whenever you push.
|
| 30 |
-
|
| 31 |
-
### B. The "CI/CD" Method (Advanced)
|
| 32 |
-
Use **GitHub Actions** to automatically upload models to Hugging Face after a successful training run in Dagster.
|
| 33 |
-
1. Generate a **Hugging Face Access Token** (Write permission).
|
| 34 |
-
2. Add it as a **Secret** in your GitHub Organization.
|
| 35 |
-
3. Add a step in your `.github/workflows/deploy.yml` to push the model artifact to Hugging Face.
|
| 36 |
-
|
| 37 |
-
## 3. Step-by-Step Setup Checklist
|
| 38 |
-
|
| 39 |
-
1. **On GitHub:**
|
| 40 |
-
- Create a GitHub Organization: [github.com/organizations/plan](https://github.com/organizations/plan).
|
| 41 |
-
- Invite your personal account as the "Owner."
|
| 42 |
-
|
| 43 |
-
2. **On Hugging Face:**
|
| 44 |
-
- Create an Organization: [huggingface.co/organizations/new](https://huggingface.co/organizations/new).
|
| 45 |
-
- Invite your personal account as the "Admin."
|
| 46 |
-
|
| 47 |
-
3. **Link the Two:**
|
| 48 |
-
- **Step 1:** Create your Space or Model repository on Hugging Face first (as a "Static" space).
|
| 49 |
-
- **Step 2:** Once created, click on the **"Settings"** tab at the top of that specific Space/Model page.
|
| 50 |
-
- **Step 3:** Scroll down to the **"GitHub Mirroring"** or **"Repository Mirroring"** section.
|
| 51 |
-
- **Step 4:** Follow the instructions there to add your GitHub repository URL and a "Mirroring Token" from GitHub.
|
| 52 |
-
- **Method 2 (The Pro Way):** Use **GitHub Actions**. This is what most startups do. It pushes your code from GitHub to Hugging Face automatically whenever you save. We will set this up together in the "CI/CD" phase.
|
| 53 |
-
|
| 54 |
-
## 4. Why this matters for the Thesis & Business
|
| 55 |
-
- **For the Thesis:** It shows you understand **Data Provenance**. You have the code on GitHub and the resulting model on Hugging Face, linked together.
|
| 56 |
-
- **For the Business:** It creates a "Hugging Face Portfolio." When someone searches for "AIOps" on Hugging Face, they will find your organization and your high-quality, documented models.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/business/platform_selection.md
DELETED
|
@@ -1,52 +0,0 @@
|
|
| 1 |
-
# AIONYX: Platform Selection & Hosting Strategy
|
| 2 |
-
|
| 3 |
-
Choosing where to host **AIONYX** is a strategic decision. You need a platform that provides academic visibility for your thesis and a professional foundation for your future business.
|
| 4 |
-
|
| 5 |
-
## 1. Platform Comparison Table
|
| 6 |
-
|
| 7 |
-
| Platform | Best For | Pros | Cons |
|
| 8 |
-
| :--- | :--- | :--- | :--- |
|
| 9 |
-
| **GitHub** | **The Default Choice** | Massive community, Industry standard for DevOps, Best CI/CD (Actions). | Public private repos have some limits for teams. |
|
| 10 |
-
| **HuggingFace** | **AI/ML Assets** | Built for ML. Best for hosting models and datasets. Very "trendy" for AIOps. | Not a full "Git" host for infra/DevOps code. |
|
| 11 |
-
| **GitLab** | **Deep DevOps** | Best built-in CI/CD pipelines, great for self-hosting. | Slightly more complex for a solo student. |
|
| 12 |
-
| **Bitbucket** | **Corporate/Jira** | Good integration with Jira/Confluence. | Lower community visibility. |
|
| 13 |
-
|
| 14 |
-
## 2. Recommendation: The "Power Duo" (GitHub + HuggingFace)
|
| 15 |
-
|
| 16 |
-
For your situation (Student + Solo Founder), I recommend using **GitHub** as your primary headquarters and **HuggingFace** as your "Model Store."
|
| 17 |
-
|
| 18 |
-
### Why GitHub for the Codebase:
|
| 19 |
-
1. **Academic Visibility**: When recruiters or examiners look at your work, they look at GitHub. It is your "Digital Resume."
|
| 20 |
-
2. **DevOps Ecosystem**: Your Terraform, K8s manifests, and Python code will thrive there because of **GitHub Actions** and the vast library of community "Actions."
|
| 21 |
-
3. **The "Startup" Path**: You can easily transition from a personal project to a company.
|
| 22 |
-
|
| 23 |
-
### Why HuggingFace for the Models:
|
| 24 |
-
1. **ML Identity**: Storing your anomaly detection models and the datasets you collect during your thesis on HuggingFace shows you are a serious Data Scientist.
|
| 25 |
-
2. **Free Hosting for Large Files**: Git is bad at storing large `.pkl` or `.pt` model files. HuggingFace is built for them.
|
| 26 |
-
|
| 27 |
-
---
|
| 28 |
-
|
| 29 |
-
## 3. GitHub Organizations: Thinking Like a Business
|
| 30 |
-
|
| 31 |
-
Since you mentioned being a student but wanting to build a business, you should create a **GitHub Organization** instead of just a personal repo.
|
| 32 |
-
|
| 33 |
-
### Why use an Organization (e.g., `github.com/aionyx-hq`)?
|
| 34 |
-
- **Professionalism**: It looks like a real entity from day one. `aionyx-hq/platform` looks better than `user123/aionyx`.
|
| 35 |
-
- **Branding**: You can set a logo, a profile, and a dedicated README for the whole group.
|
| 36 |
-
- **Permission Management**: Today you are alone. Tomorrow when you hire your first intern, you can give them access to *some* repos without giving them your personal account password.
|
| 37 |
-
- **GitHub Pages**: You can host your project documentation (`aionyx.github.io`) beautifully.
|
| 38 |
-
|
| 39 |
-
### Is it Free for Students?
|
| 40 |
-
**Yes.** GitHub Organizations are free for unlimited public and private repositories for teams. As a student, you also get the **GitHub Student Developer Pack**, which gives you extra "Pro" features for free for your organization.
|
| 41 |
-
|
| 42 |
-
---
|
| 43 |
-
|
| 44 |
-
## 4. The Execution Plan
|
| 45 |
-
|
| 46 |
-
1. **Create an Organization**: Name it something like `aionyx-ops`, `aionyx-ai`, or `aionyx-hq`.
|
| 47 |
-
2. **Create Private Repos (Optional)**: While you work on your thesis, you can keep some "secret sauce" private.
|
| 48 |
-
3. **Public "Lead Magnet"**: Once the thesis is submitted, make the main orchestration logic public to attract the community.
|
| 49 |
-
4. **Connect HuggingFace**: Create a HuggingFace account for `aionyx` to host your trained models.
|
| 50 |
-
|
| 51 |
-
---
|
| 52 |
-
**Verdict**: Go with a **GitHub Organization**. It’s the most professional, free, and scalable option for a student transitioning into a startup founder.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/thesis/chapter_mapping.md
DELETED
|
@@ -1,25 +0,0 @@
|
|
| 1 |
-
# AIONYX Thesis Alignment Roadmap
|
| 2 |
-
|
| 3 |
-
As your professor, here is how the repository structure and toolset map directly to your **Master's Thesis chapters** at Malmö University.
|
| 4 |
-
|
| 5 |
-
| Thesis Chapter | Repository Section | Tool/Tech Focus |
|
| 6 |
-
| :--- | :--- | :--- |
|
| 7 |
-
| **1. Introduction** | `docs/thesis/proposal.md` | Problem statement, research questions. |
|
| 8 |
-
| **2. Background** | `docs/architecture/` | Kubernetes (K8s), AIOps state-of-the-art. |
|
| 9 |
-
| **3. Methodology** | `platform/orchestration/` | **Dagster** for reproducible data pipelines. |
|
| 10 |
-
| **4. System Design** | `infra/` + `platform/storage/` | **Terraform**, **K8s**, **Postgres** schema. |
|
| 11 |
-
| **5. Implementation** | `platform/ml/` | **Python**, **Scikit-learn/PyTorch**. |
|
| 12 |
-
| **6. Experiments** | `experiments/notebooks/` | Jupyter, Dataset analysis, Model evaluation. |
|
| 13 |
-
| **7. Evaluation** | `docs/thesis/results.md` | Metrics (Precision, Recall, Cost savings). |
|
| 14 |
-
| **8. Conclusion** | `docs/business/roadmap.md` | Future outlook (Thesis to Startup transition). |
|
| 15 |
-
|
| 16 |
-
## Academic Contribution Checklist
|
| 17 |
-
|
| 18 |
-
- [ ] **Reproducibility**: Guaranteed by the `Dagster` asset definitions and `Terraform` manifests.
|
| 19 |
-
- [ ] **Data Quality**: Managed through `Dagster` software-defined assets and type checking.
|
| 20 |
-
- [ ] **Novelty**: Applying ML particularly for *proactive* failure detection rather than reactive monitoring.
|
| 21 |
-
- [ ] **Validity**: Experimental results backed by the `model_registry` in Postgres.
|
| 22 |
-
|
| 23 |
-
## Professor's Tip for the Viva (Defense)
|
| 24 |
-
When asked about your choice of **Dagster**, explain:
|
| 25 |
-
> "I chose Dagster instead of simple scripts because it provides a self-documenting, software-defined asset graph. This ensures that the data used for the thesis results is traceable, reproducible, and verifiable—standard requirements for high-quality scientific research."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/thesis/literature_review.md
DELETED
|
@@ -1,58 +0,0 @@
|
|
| 1 |
-
# Literature Review: AIOps, Predictive Reliability, and Kubernetes
|
| 2 |
-
|
| 3 |
-
This document serves as the academic foundation for the **AIONYX** thesis. It outlines current research trends, key terminology, and a guide for further study.
|
| 4 |
-
|
| 5 |
-
## 1. Key Research Areas (What to Study)
|
| 6 |
-
|
| 7 |
-
To write a high-scoring thesis at Malmö University, you must demonstrate a deep understanding of these four pillars:
|
| 8 |
-
|
| 9 |
-
### A. AIOps (Artificial Intelligence for IT Operations)
|
| 10 |
-
Focus on the transition from "Monitoring" (knowing what happened) to "AIOps" (predicting what will happen).
|
| 11 |
-
- **Core Concept:** Multi-source data ingestion (metrics, logs, events) + ML for automated remediation.
|
| 12 |
-
- **Key Terms:** MTTR (Mean Time To Recovery), MTBF (Mean Time Between Failures), Noise Reduction, Alert Correlation.
|
| 13 |
-
|
| 14 |
-
### B. Time-Series Anomaly Detection
|
| 15 |
-
Since Kubernetes telemetry is time-sequential, you must study how to detect "out-of-pattern" behavior.
|
| 16 |
-
- **Algorithms:** LSTMs (Long Short-Term Memory), Isolation Forests, Autoencoders, and Prophet (by Meta).
|
| 17 |
-
- **Sub-topic:** "Grey Failures"—failures that don't trigger a binary health check but degrade performance.
|
| 18 |
-
|
| 19 |
-
### C. Predictive Infrastructure Scaling
|
| 20 |
-
This links your project to **Cost Optimization**.
|
| 21 |
-
- **Research focus:** Horizontal Pod Autoscaler (HPA) vs. Predictive Scaling (using ML to scale *before* a spike).
|
| 22 |
-
- **Concept:** Trade-offs between "Resource Under-utilization" (waste) and "Service Level Agreement (SLA) Violations" (risk).
|
| 23 |
-
|
| 24 |
-
### D. Reproducible MLOps in SRE
|
| 25 |
-
This is where **Dagster** fits into the research.
|
| 26 |
-
- **Focus:** How "Software-Defined Assets" improve the reliability of ML models in production infrastructure.
|
| 27 |
-
- **Key Term:** Data Provenance (knowing exactly where your training data came from).
|
| 28 |
-
|
| 29 |
-
---
|
| 30 |
-
|
| 31 |
-
## 2. Recommended Search Strategies (Where to Look)
|
| 32 |
-
|
| 33 |
-
Use these platforms for peer-reviewed, high-impact papers:
|
| 34 |
-
|
| 35 |
-
1. **IEEE Xplore / ACM Digital Library:** Search for *"Anomalous behavior detection in Kubernetes"* or *"Deep learning for microservices reliability."*
|
| 36 |
-
2. **Google Scholar:** Use specific queries like: `"AIOps" + "Kubernetes" + "Predictive maintenance"`.
|
| 37 |
-
3. **ArXiv.org (Computer Science section):** For the most cutting-edge (but sometimes unreviewed) papers on AI and Systems.
|
| 38 |
-
4. **Malmö University Library (LibSearch):** Specifically look for previous theses in **Data Science** or **Computer Science** related to "Cloud Computing" or "Monitoring."
|
| 39 |
-
|
| 40 |
-
---
|
| 41 |
-
|
| 42 |
-
## 3. High-Value Papers to Find (Search Keywords)
|
| 43 |
-
|
| 44 |
-
Look for these specific topics to build your bibliography:
|
| 45 |
-
- *"MicroRCA: Root Cause Analysis of Microservices in Clusters"* (Very famous paper).
|
| 46 |
-
- *"Cloud-native system anomaly detection using deep learning."*
|
| 47 |
-
- *"Cost-aware resource management in containerized environments."*
|
| 48 |
-
- *The "Google SRE Book"* (Specifically the chapters on Monitoring and Alerting—this is the "Bible" of DevOps).
|
| 49 |
-
|
| 50 |
-
---
|
| 51 |
-
|
| 52 |
-
## 4. How to Read for the Thesis
|
| 53 |
-
|
| 54 |
-
When reading, don't just summarize. **Compare.**
|
| 55 |
-
- "Paper A uses LSTMs, but Paper B argues that Isolation Forests are faster for real-time K8s metrics. **AIONYX** will test both to find the best balance for small-scale startups."
|
| 56 |
-
|
| 57 |
-
> [!TIP]
|
| 58 |
-
> Use a tool like **Zotero** or **Mendeley** (both free) to save your papers and automatically generate citations. This will save you weeks of work during the writing phase.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
docs/thesis/proposal.md
DELETED
|
@@ -1,53 +0,0 @@
|
|
| 1 |
-
# AIONYX: Project Proposal (Thesis + Business)
|
| 2 |
-
|
| 3 |
-
## 1. Executive Summary
|
| 4 |
-
**AIONYX** is a data-driven AIOps platform designed to provide predictive reliability and cost efficiency for Kubernetes environments. It bridges the gap between infrastructure automation and machine learning, offering a system that moves beyond reactive monitoring into proactive operations.
|
| 5 |
-
|
| 6 |
-
## 2. The Vision: "Lean & Mean" AIOps
|
| 7 |
-
As a one-person project on a student budget, AIONYX focuses on a **Zero-Cost, High-Impact** stack. We use the best open-source tools to build a system that rivals expensive commercial alternatives (Datadog, Dynatrace).
|
| 8 |
-
|
| 9 |
-
### Core Research Question (Thesis)
|
| 10 |
-
*How can a software-defined ingestion pipeline (Dagster) and machine learning models (AIOps) be used to predict Kubernetes node failures 30 minutes before they occur?*
|
| 11 |
-
|
| 12 |
-
## 3. Implementation Strategy (The "Solo" Plan)
|
| 13 |
-
To succeed as a single developer, the project is scoped around a single "Killer Feature": **Predictive Anomaly Detection**.
|
| 14 |
-
|
| 15 |
-
### Phase 1: Local Foundation (Month 1-2)
|
| 16 |
-
- **Environment**: Use `Kind` or `Minikube` (Free) instead of expensive cloud clusters.
|
| 17 |
-
- **Ingestion**: Dagster assets pulling `container_cpu_usage_seconds_total` and `container_memory_working_set_bytes`.
|
| 18 |
-
- **Storage**: Postgres (Standard) or TimescaleDB (vial Docker - Free).
|
| 19 |
-
|
| 20 |
-
### Phase 2: Data & ML (Month 3-4)
|
| 21 |
-
- **Feature Engineering**: Transform raw metrics into time-series features (Rolling mean, Std dev, Rate of change).
|
| 22 |
-
- **Model**: Start with **Isolation Forest** (Unsupervised) to detect outliers, then move to **LSTMs** if time permits.
|
| 23 |
-
- **Labels**: Use your internship experience to "inject" failures (Stress tests) to create a labeled dataset for free.
|
| 24 |
-
|
| 25 |
-
### Phase 3: Business MVP (Month 5-6)
|
| 26 |
-
- **Alerting**: Simple Slack/Discord integration (Free).
|
| 27 |
-
- **Dashboard**: Use Dagster's built-in UI (Dagit) as your first "product" interface—no custom frontend needed yet.
|
| 28 |
-
|
| 29 |
-
## 4. Zero-Cost Market Standard Stack
|
| 30 |
-
| Function | Tool | Commercial Equivalent | Cost |
|
| 31 |
-
| :--- | :--- | :--- | :--- |
|
| 32 |
-
| **Orchestration** | Dagster | Airflow/Prefect | $0 |
|
| 33 |
-
| **Database** | Postgres | Snowflake/Databricks | $0 |
|
| 34 |
-
| **Compute** | Local K8s (Kind) | EKS/GKE | $0 |
|
| 35 |
-
| **CI/CD** | GitHub Actions | Jenkins/CircleCI | $0 ($2k free mins) |
|
| 36 |
-
| **Infrastructure** | Terraform | Manual Config | $0 |
|
| 37 |
-
| **ML Framework** | Scikit-learn/PyTorch| SageMaker | $0 |
|
| 38 |
-
|
| 39 |
-
## 5. Business Path: From Thesis to Startup
|
| 40 |
-
1. **Open Source (The "Lead Magnet")**: Release the core Dagster assets as a public repo. This builds your brand name in the DevOps community.
|
| 41 |
-
2. **Productize**: Offer a "Cloud Agent" that small startups can drop into their clusters to get instant ML insights without setting up Dagster themselves.
|
| 42 |
-
3. **Target Market**: Startups with < 50 nodes who cannot afford $5,000/month Datadog bills.
|
| 43 |
-
|
| 44 |
-
## 6. Milestones & Deliverables
|
| 45 |
-
- **M1**: Reproducible local K8s + Dagster + Postgres environment (Infrastructure).
|
| 46 |
-
- **M2**: Automated data ingestion and "Feature Store" populated with simulated failures.
|
| 47 |
-
- **M3**: Trained ML model with > 80% precision on failure prediction.
|
| 48 |
-
- **M4**: Final Thesis at Malmö University + Open Source MVP launch.
|
| 49 |
-
|
| 50 |
-
---
|
| 51 |
-
**Prepared by**: [User Name]
|
| 52 |
-
**Academic Advisor**: Your AI Professor
|
| 53 |
-
**Date**: December 2025
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
infra/terraform/kind/aionyx-cluster-config
DELETED
|
@@ -1,19 +0,0 @@
|
|
| 1 |
-
apiVersion: v1
|
| 2 |
-
clusters:
|
| 3 |
-
- cluster:
|
| 4 |
-
certificate-authority-data: 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
|
| 5 |
-
server: https://127.0.0.1:55539
|
| 6 |
-
name: kind-aionyx-cluster
|
| 7 |
-
contexts:
|
| 8 |
-
- context:
|
| 9 |
-
cluster: kind-aionyx-cluster
|
| 10 |
-
user: kind-aionyx-cluster
|
| 11 |
-
name: kind-aionyx-cluster
|
| 12 |
-
current-context: kind-aionyx-cluster
|
| 13 |
-
kind: Config
|
| 14 |
-
preferences: {}
|
| 15 |
-
users:
|
| 16 |
-
- name: kind-aionyx-cluster
|
| 17 |
-
user:
|
| 18 |
-
client-certificate-data: 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
|
| 19 |
-
client-key-data: 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
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
infra/terraform/kind/main.tf
DELETED
|
@@ -1,56 +0,0 @@
|
|
| 1 |
-
terraform {
|
| 2 |
-
required_providers {
|
| 3 |
-
kind = {
|
| 4 |
-
source = "tehcyx/kind"
|
| 5 |
-
version = "0.2.1"
|
| 6 |
-
}
|
| 7 |
-
kubernetes = {
|
| 8 |
-
source = "hashicorp/kubernetes"
|
| 9 |
-
version = "2.23.0"
|
| 10 |
-
}
|
| 11 |
-
helm = {
|
| 12 |
-
source = "hashicorp/helm"
|
| 13 |
-
version = "2.11.0"
|
| 14 |
-
}
|
| 15 |
-
}
|
| 16 |
-
}
|
| 17 |
-
|
| 18 |
-
provider "kind" {}
|
| 19 |
-
|
| 20 |
-
provider "helm" {
|
| 21 |
-
kubernetes {
|
| 22 |
-
host = kind_cluster.aionyx_cluster.endpoint
|
| 23 |
-
client_certificate = kind_cluster.aionyx_cluster.client_certificate
|
| 24 |
-
client_key = kind_cluster.aionyx_cluster.client_key
|
| 25 |
-
cluster_ca_certificate = kind_cluster.aionyx_cluster.cluster_ca_certificate
|
| 26 |
-
}
|
| 27 |
-
}
|
| 28 |
-
|
| 29 |
-
resource "kind_cluster" "aionyx_cluster" {
|
| 30 |
-
name = "aionyx-cluster"
|
| 31 |
-
wait_for_ready = true
|
| 32 |
-
|
| 33 |
-
kind_config {
|
| 34 |
-
kind = "Cluster"
|
| 35 |
-
api_version = "kind.x-k8s.io/v1alpha4"
|
| 36 |
-
|
| 37 |
-
node {
|
| 38 |
-
role = "control-plane"
|
| 39 |
-
}
|
| 40 |
-
|
| 41 |
-
node {
|
| 42 |
-
role = "worker"
|
| 43 |
-
}
|
| 44 |
-
|
| 45 |
-
node {
|
| 46 |
-
role = "worker"
|
| 47 |
-
}
|
| 48 |
-
}
|
| 49 |
-
}
|
| 50 |
-
|
| 51 |
-
provider "kubernetes" {
|
| 52 |
-
host = kind_cluster.aionyx_cluster.endpoint
|
| 53 |
-
client_certificate = kind_cluster.aionyx_cluster.client_certificate
|
| 54 |
-
client_key = kind_cluster.aionyx_cluster.client_key
|
| 55 |
-
cluster_ca_certificate = kind_cluster.aionyx_cluster.cluster_ca_certificate
|
| 56 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
infra/terraform/kind/observability.tf
DELETED
|
@@ -1,26 +0,0 @@
|
|
| 1 |
-
resource "kubernetes_namespace" "monitoring" {
|
| 2 |
-
metadata {
|
| 3 |
-
name = "monitoring"
|
| 4 |
-
}
|
| 5 |
-
}
|
| 6 |
-
|
| 7 |
-
resource "helm_release" "prometheus" {
|
| 8 |
-
name = "prometheus"
|
| 9 |
-
repository = "https://prometheus-community.github.io/helm-charts"
|
| 10 |
-
chart = "kube-prometheus-stack"
|
| 11 |
-
namespace = kubernetes_namespace.monitoring.metadata[0].name
|
| 12 |
-
version = "51.2.0"
|
| 13 |
-
|
| 14 |
-
set {
|
| 15 |
-
name = "grafana.enabled"
|
| 16 |
-
value = "true"
|
| 17 |
-
}
|
| 18 |
-
|
| 19 |
-
set {
|
| 20 |
-
name = "prometheus.prometheusSpec.serviceMonitorSelectorNilUsesHelmValues"
|
| 21 |
-
value = "false"
|
| 22 |
-
}
|
| 23 |
-
|
| 24 |
-
# Ensure the cluster is ready before deploying
|
| 25 |
-
depends_on = [kind_cluster.aionyx_cluster]
|
| 26 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
infra/terraform/kind/postgres.tf
DELETED
|
@@ -1,36 +0,0 @@
|
|
| 1 |
-
resource "kubernetes_namespace" "storage" {
|
| 2 |
-
metadata {
|
| 3 |
-
name = "storage"
|
| 4 |
-
}
|
| 5 |
-
}
|
| 6 |
-
|
| 7 |
-
resource "helm_release" "postgres" {
|
| 8 |
-
name = "aionyx-db"
|
| 9 |
-
repository = "https://charts.bitnami.com/bitnami"
|
| 10 |
-
chart = "postgresql"
|
| 11 |
-
namespace = kubernetes_namespace.storage.metadata[0].name
|
| 12 |
-
version = "12.12.10"
|
| 13 |
-
|
| 14 |
-
set {
|
| 15 |
-
name = "auth.username"
|
| 16 |
-
value = "aionyx"
|
| 17 |
-
}
|
| 18 |
-
|
| 19 |
-
set {
|
| 20 |
-
name = "auth.password"
|
| 21 |
-
value = "aionyx_secret_123"
|
| 22 |
-
}
|
| 23 |
-
|
| 24 |
-
set {
|
| 25 |
-
name = "auth.database"
|
| 26 |
-
value = "aionyx_telemetry"
|
| 27 |
-
}
|
| 28 |
-
|
| 29 |
-
set {
|
| 30 |
-
name = "primary.persistence.size"
|
| 31 |
-
value = "10Gi"
|
| 32 |
-
}
|
| 33 |
-
|
| 34 |
-
# Ensure the cluster is ready
|
| 35 |
-
depends_on = [kind_cluster.aionyx_cluster]
|
| 36 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
infra/terraform/kind/tfplan_obs
DELETED
|
Binary file (47.2 kB)
|
|
|
platform/orchestration/dagster/__init__.py
DELETED
|
File without changes
|
platform/orchestration/dagster/assets/__init__.py
DELETED
|
File without changes
|
platform/orchestration/dagster/assets/ingestion.py
DELETED
|
@@ -1,31 +0,0 @@
|
|
| 1 |
-
from dagster import asset, ResourceParam
|
| 2 |
-
import pandas as pd
|
| 3 |
-
from datetime import datetime
|
| 4 |
-
|
| 5 |
-
@asset(group_name="ingestion", compute_kind="python")
|
| 6 |
-
def raw_kubernetes_metrics() -> pd.DataFrame:
|
| 7 |
-
"""
|
| 8 |
-
Mock asset representing raw metrics pulled from Prometheus/Kubernetes.
|
| 9 |
-
In production, this would use a Prometheus API client.
|
| 10 |
-
"""
|
| 11 |
-
# Simulate some data for now
|
| 12 |
-
data = {
|
| 13 |
-
"timestamp": [datetime.now() for _ in range(5)],
|
| 14 |
-
"node_name": ["node-1", "node-2", "node-1", "node-3", "node-2"],
|
| 15 |
-
"metric_name": ["cpu_usage", "memory_usage", "cpu_usage", "cpu_usage", "memory_usage"],
|
| 16 |
-
"value": [0.45, 0.78, 0.42, 0.91, 0.82]
|
| 17 |
-
}
|
| 18 |
-
return pd.DataFrame(data)
|
| 19 |
-
|
| 20 |
-
@asset(group_name="ingestion", compute_kind="python")
|
| 21 |
-
def raw_kubernetes_logs() -> pd.DataFrame:
|
| 22 |
-
"""
|
| 23 |
-
Mock asset representing logs pulled from Kubernetes API or ELK.
|
| 24 |
-
"""
|
| 25 |
-
data = {
|
| 26 |
-
"timestamp": [datetime.now() for _ in range(3)],
|
| 27 |
-
"pod_name": ["api-pod", "db-pod", "auth-pod"],
|
| 28 |
-
"log_level": ["INFO", "ERROR", "INFO"],
|
| 29 |
-
"message": ["Server started", "Connection failed", "User logged in"]
|
| 30 |
-
}
|
| 31 |
-
return pd.DataFrame(data)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
platform/orchestration/dagster/assets/ml_models.py
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
from dagster import asset
|
| 2 |
-
import pandas as pd
|
| 3 |
-
|
| 4 |
-
@asset(deps=["anomaly_labels"], group_name="ml")
|
| 5 |
-
def trained_anomaly_model(anomaly_labels: pd.DataFrame):
|
| 6 |
-
"""
|
| 7 |
-
Mock asset representing the training of an ML model.
|
| 8 |
-
In production, this would use scikit-learn/PyTorch and save to model_registry.
|
| 9 |
-
"""
|
| 10 |
-
# Simulate training logic
|
| 11 |
-
model_info = {
|
| 12 |
-
"model_type": "IsolationForest",
|
| 13 |
-
"precision": 0.85,
|
| 14 |
-
"recall": 0.72,
|
| 15 |
-
"features_used": ["cpu_usage", "rolling_mean"]
|
| 16 |
-
}
|
| 17 |
-
return model_info
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
platform/orchestration/dagster/assets/processing.py
DELETED
|
@@ -1,23 +0,0 @@
|
|
| 1 |
-
from dagster import asset
|
| 2 |
-
import pandas as pd
|
| 3 |
-
|
| 4 |
-
@asset(deps=["raw_kubernetes_metrics"], group_name="processing")
|
| 5 |
-
def feature_cpu_usage(raw_kubernetes_metrics: pd.DataFrame) -> pd.DataFrame:
|
| 6 |
-
"""
|
| 7 |
-
Transform raw metrics into features.
|
| 8 |
-
In a real scenario, this would compute rolling averages,
|
| 9 |
-
detect spikes, or perform normalization.
|
| 10 |
-
"""
|
| 11 |
-
df = raw_kubernetes_metrics[raw_kubernetes_metrics["metric_name"] == "cpu_usage"].copy()
|
| 12 |
-
# Mock feature engineering: normalized value and rolling mean
|
| 13 |
-
df["rolling_mean"] = df["value"].rolling(window=2, min_periods=1).mean()
|
| 14 |
-
return df
|
| 15 |
-
|
| 16 |
-
@asset(deps=["feature_cpu_usage"], group_name="processing")
|
| 17 |
-
def anomaly_labels(feature_cpu_usage: pd.DataFrame) -> pd.DataFrame:
|
| 18 |
-
"""
|
| 19 |
-
Simulate manual or automated labeling for supervised learning.
|
| 20 |
-
"""
|
| 21 |
-
df = feature_cpu_usage.copy()
|
| 22 |
-
df["label"] = df["value"].apply(lambda x: "anomaly" if x > 0.9 else "healthy")
|
| 23 |
-
return df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
platform/orchestration/dagster/definitions.py
DELETED
|
@@ -1,11 +0,0 @@
|
|
| 1 |
-
from dagster import Definitions, load_assets_from_modules
|
| 2 |
-
from .assets import ingestion, processing, ml_models
|
| 3 |
-
|
| 4 |
-
all_assets = load_assets_from_modules([ingestion, processing, ml_models])
|
| 5 |
-
|
| 6 |
-
defs = Definitions(
|
| 7 |
-
assets=all_assets,
|
| 8 |
-
# resources={
|
| 9 |
-
# "postgres": postgres_resource,
|
| 10 |
-
# },
|
| 11 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|