Notes
f4: Facebook's Warm BLOB Storage System
Facebook's f4 is a warm BLOB storage system that reduces storage overhead by using erasure coding instead of replication while maintaining high availability and reliability for aged data.
Finding a needle in Haystack: Facebook’s photo storage
Facebook's Haystack is an object storage system optimized for the massive scale of photo storage, reducing disk IO by keeping metadata in memory.
Facebook’s Tectonic Filesystem: Efficiency from Exascale
Tectonic is Facebook’s exabyte-scale distributed filesystem that consolidates large tenants into general multitenant instances for efficiency.
Building and Operating a Pretty Big Storage System Called S3 (Video)
Andy Warfield shares his five years of experience building Amazon S3, covering hard drive fleet management, durability practices, and organizational scale.
S3 Deep Dive: Building and Operating for Resiliency
AWS engineers explain S3's threat modeling culture, 11 nines durability design, and availability architecture across availability zones.
2022 Re:Invent, Deep Dive on Amazon S3: Fundamentals of Durability, Performance, Cost, and Security
AWS re:Invent session covering Amazon S3's four core fundamentals: durability via erasure coding and repair mechanisms, horizontal scaling for performance, storage class optimization for cost, and IAM-based security controls.
Building and Operating a Pretty Big Storage System Called S3 (Blog)
Andy Warfield shares insights from six years working on Amazon S3, covering technical challenges of operating at massive scale, heat management across millions of hard drives, and the human factors of building reliable systems.
WhatsApp Web's Data Journey: Storage Locations and End-to-End Encryption
An exploration of how WhatsApp Web manages local data storage using IndexedDB and maintains end-to-end encryption security.
Writing a good CLAUDE.md
A guide on crafting effective CLAUDE.md files to provide context and instructions for LLM agents.
Zuul 2: The Netflix Journey to Asynchronous, Non-Blocking Systems
Netflix's journey migrating their cloud gateway, Zuul, from a blocking to an asynchronous, non-blocking architecture to handle massive scale.
How Discord Stores Trillions of Messages
Discord migrated from Cassandra to ScyllaDB and implemented Rust-based intermediate data services to handle trillions of messages with low latency.
Coordination
Explores thread coordination mechanisms including shared state (blocking queues) and message passing (actor model), with practical patterns for async processing and bursty traffic.
Correctness in Concurrency: Low Level Design
A deep dive into ensuring correctness in concurrent systems through proper locking strategies, atomic operations, and avoiding common pitfalls like race conditions and deadlocks.
Design Principles for Low-Level Design Interviews
A comprehensive guide to effectively applying general and object-oriented design principles like KISS, DRY, and SOLID in low-level design interviews.
Low-Level Design Delivery Framework
A step-by-step framework for structuring low-level design interviews, covering requirements, entities, class design, implementation, and extensibility to ensure pacing and meaningful progress.
OOP Concepts
A focused refresher on core OOP concepts—Encapsulation, Abstraction, Polymorphism, and Inheritance—tailored for Low-Level Design interviews.
Low-Level Design Patterns
A focused guide to the essential design patterns for low-level design interviews, distinguishing between historical GoF patterns and what actually appears in modern interviews.
API Design
A comprehensive guide to designing APIs for system design interviews, covering protocols, resource modeling, and common patterns.
C++ Threading and Concurrency
A comprehensive crash course on C++ threading and concurrency.
Data Modeling
A practical guide to data modeling for system design interviews, focusing on database selection, schema design key drivers, and scaling strategies.
Database Indexing
A deep dive into database indexing, covering how indexes work, their costs and trade-offs, and a detailed look at B-Tree indexes.
How Not to Bomb Your Offer Negotiation
A guide to negotiating job offers with a collaborative mindset, covering how to handle salary expectations and counteroffers.
Ten Rules for Negotiating a Job Offer
A practical guide to job offer negotiation, focusing on mindset, leverage, and strategic communication.
Networking Essentials for System Design
A comprehensive guide to networking concepts for system design, covering layers, protocols (TCP/UDP, HTTP, gRPC, etc.), load balancing, and failure handling.
OpenSpec Deep Dive: Spec-Driven Development Architecture & Practice
A comprehensive guide to OpenSpec, a spec-driven development architecture designed to enhance AI-assisted programming workflows through structured constraints and context management.
OpenSpec vs Spec Kit: Choosing the Right AI-Driven Development Workflow for Your Team
A comparison of OpenSpec and Spec Kit workflows, highlighting OpenSpec's speed and conciseness versus Spec Kit's structured approach.
System Design Core Concepts: Caching
A comprehensive guide to caching strategies, patterns, and best practices for system design interviews, covering architectures, eviction policies, and common pitfalls.
Levels.fyi's Over-The-Air Mobile Updates
How Levels.fyi built a cost-effective, self-hosted serverless OTA update system to replace Microsoft AppCenter.
Optimizing Our React Performance
Strategies for optimizing React performance, including avoiding unnecessary re-renders and handling large lists.
How Levels.fyi scaled to millions of users with Google Sheets as a backend
Levels.fyi scaled to millions of users without a traditional backend by leveraging Google Sheets, Google Forms, and AWS Lambda.
Reimagining LinkedIn’s search tech stack
A deep dive into how LinkedIn rebuilt its search stack.
Effective harnesses for long-running agents
A framework for long-running agents using initializer and coding agents to maintain context and ensure incremental progress.
How we built Pingora, the proxy that connects Cloudflare to the Internet
Cloudflare built Pingora, a new HTTP proxy in Rust, to replace NGINX, serving over 1 trillion requests daily with improved performance and efficiency.