đ° Engineering Blogs & Papers
Blogs (read 2 posts/week and write a 5-line summary in Notes/)
- â AWS Buildersâ Library: timeouts/retries/backoff, shuffle sharding, static stability
- â Marc Brookerâs blog: distributed systems at AWS
- Uber, Netflix, Discord (âHow Discord stores trillions of messagesâ), Stripe (idempotency, rate limiters), Cloudflare, Shopify, Figma (multiplayer), Slack, Airbnb, LinkedIn, Meta, Dropbox, Pinterest
- Indian tech: Flipkart, Swiggy Bytes, Zomato, Razorpay, PhonePe, Zerodha Tech
- High Scalability archive · The Architecture of Open Source Applications (free)
- Martin Fowler (martinfowler.com) · Jepsen analyses · Brendan Gregg · Dan Luu
Papers (read one every 1â2 weeks from Phase 3; use the âthree-passâ method)
| Paper | Why |
|---|---|
| â In Search of an Understandable Consensus Algorithm (Raft) | Consensus, MIT lab |
| â Dynamo: Amazonâs Highly Available Key-value Store | Leaderless, quorums, vector clocks |
| MapReduce | Batch processing, MIT lab 1 |
| The Google File System | Distributed storage |
| Bigtable | Wide-column, LSM |
| Spanner | Global consistency, TrueTime |
| Kafka: a Distributed Messaging System for Log Processing | The log |
| The Chubby lock service | Coordination |
| Zanzibar: Googleâs Consistent, Global Authorization System | ReBAC authorization |
| Time, Clocks, and the Ordering of Events (Lamport) | Logical time |
| â Attention Is All You Need | Transformers |
| Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks | RAG |
| ReAct: Synergizing Reasoning and Acting in Language Models | Agents |
| On Designing and Deploying Internet-Scale Services (Hamilton) | Ops wisdom |
How to read a paper: S. Keshav, How to Read a Paper (three-pass approach).