All blogs / Distributed systems / cloud
Arpit Bhayani
I document my learnings and views every week, and cover various topics that interest me. Here are all the blogs I wrote to date.
Distributed systems / cloudDatabasesAI / MLAlgorithms / data structures
- Posts
- 128 from 12 Apr 2016 to 24 Oct 2026
- Posting
- weekly (43 in the last year)
- Kind
- personal blog
- Feed
- https://arpitbhayani.me/rss.xml
2026 35
- Compiling a Workflow into a 3B Model
- PostgreSQL 19 Ships Property Graphs Enabling Direct Query of Existing Tables
- Quantization Fundamentals: How LLMs Run Efficiently with Reduced Memory and Fast Arithmetic
- Reorganizing KV cache memory can make models interactive in real-time
- Engineering Speed And Abstractions Create Longer Failure Paths, An Unacknowledged Tradeoff
- Organizing a Model's KV Cache Unlocks New Interactive Abilities
- Demystifying Kafka's Exactly-Once Semantics: Idempotence, Transactions, and Fencing
- AI makes pull requests less central in open-source contribution
- Cloudflare Implements Cache Transcoding to Optimize Cache Storage with Zstandard
- Consistent Hashing: Demystifying the Algorithm and Its Simple Implementation
- Choose a Depth, Not a Duration, To Navigate Rabbit Holes
- How Google Manages Billions of Lines of Code in a Single Monorepo
- Bitcask Explained: A Log-Structured, Fast Key-Value Store
- Design Multi-Context Processing Tools as Capabilities, Not Endpoints
- Deckrun: A Modern Presentation Utility for Markdown and Self-contained HTML
- Slop Debt
- What AI First Engineering Orgs Look Like
- Three Claude Skills I Think Every Org Should Have
- G-Eval, Explained
- HyperLogLog: Efficient Cardinality Estimation with Probabilistic Data Structures
- How Redis Implements Strings: A Deep Dive into SDS and Encoding
- AI Workflows Need Topological Sort
- Embedding Models Make Or Break Your Ai App
- Temporal Primer - Building Long-Running Systems
- What Matters in Production RAG
- Structure of Every LLM Chat
- How LLMs Really Work
- Your Monolith Is Already A Distributed System
- The Curse of Dimensionality
- Databases Were Not Designed For This
- Bm25
- JOIN Algorithms
- Why Half Your Skills Expire Every Few Years
- Multi-Paxos - Consensus in Distributed Databases
- MySQL Replication Internals
2025 19
- Bloom Filters
- Clock Synchronization Is a Nightmare
- When You Increase Kafka Partitions
- Product Quantization
- The Q, K, V Matrices
- How LLM Inference Works
- What are Blocking Queues and Why We Need Them
- Heartbeats in Distributed Systems
- How Writes Work in Apache Cassandra
- Redis Replication Internals
- How Does a CDN Handle Content Replication
- Why gRPC Uses HTTP2
- Why Distributed Systems Need Consensus Algorithms Like Raft
- Why Do Databases Deadlock and How Do They Resolve It
- Why and How Cache Locality Can Make Your Code Faster
- Why Eventual Consistency is Preferred in Distributed Systems
- Why does DNS use both UDP and TCP
- How to Write Effective OKRs
- Theta sketches: set algebra over massive streams
2024 6
- When to Change Jobs - The 3P Framework
- Paper Notes - On-demand Container Loading in AWS Lambda
- Paper Notes - SQL Has Problems. We Can Fix Them Pipe Syntax In SQL
- Paper Notes - NanoLog - A Nanosecond Scale Logging System
- Paper Notes - WTF - The Who to Follow Service at Twitter
- Turn System Design Interviews into Discussions
2023 1
2022 7
- Solving the Knapsack Problem with Evolutionary Algorithms
- Generating Pseudorandom Numbers with LFSR
- Local vs Global Indexes in Partitioned Databases
- Partitioning Data - Range, Hash, and When to Use Them
- Partitioning Strategies for Distributed Databases
- Leaderless Replication
- Multi-Master Conflicts - How to Handle Them
2021 21
- Multi-Master Replication - Why Conflicts Happen
- Multi-Master Replication - Scaling Writes Across Geographies
- Monotonic Reads - How Asynchronous Replication Creates Wormholes
- Read-Your-Writes Consistency
- What Happens When a Master or Replica Fails
- Replication Formats - Statement vs Row-based
- Replication Strategies - Synchronous, Asynchronous, and Semi-Synchronous
- Master-Replica Replication
- Decoding Durability - The D in ACID
- Decoding Isolation - The I in ACID
- Decoding Atomicity - The A in ACID
- Decoding Consistency - The C in ACID
- Architectural Patterns for Distributed Systems
- Fallacies of Distributed Computing Every Engineer Should Know
- Fork Bomb
- Python Internals - Exploring Chained Comparison Operators
- Modeling Udemy's Categories in a Relational Database
- I Changed the Rules for the Python's Walrus Operator
- Fully Persistent Arrays
- How Persistent Data Structures Work
- Constant Folding in Python
2020 35
- How Python Uses String Interning to Keep Runtime Efficient
- Visualizing Recursion in Python with Just a Decorator
- Count Distinct With Minimal Memory - Flajolet Martin Algorithm
- The 2Q Algorithm - Addressing LRU's Sub-Optimality
- Israeli Queues
- 1D Terrain Generation
- MinHash - Fast Jaccard Similarity at Scale
- Time Series Smoothing for Anomaly Detection
- Implementing LFU in O(1)
- Space-Efficient Counting - Exploring Morris' Algorithm
- Why Slowsort is Hilariously Inefficient
- Bitcask - A Log-Structured fast KV store
- Phi Accrual Failure Detection Algorithm
- Breaking Repeating-Key XOR Encryption
- ETAOIN SHRDLU - Using Letter Frequency to Decipher the Ciphered
- Making Integers Iterable in Python
- Structure Composition and Emulating Inheritance in C
- The RUM Conjecture - Storage System Trade-offs
- Consistent Hashing - What It Is and How to Implement It
- How Python Handles Integers Under the Hood
- Fractional Cascading - Speeding Up Multiple Binary Searches
- Copy-On-Write - When to Use It, When to Avoid It
- How MySQL Avoids Performance Hits from Table Scans
- Modeling Finite State Machines with Python Coroutines
- Building Better Ratings with Bayesian Averages
- Sliding Window Rate Limiting - Design and Implementation
- The Intuition Behind IDF
- Habits that Make a Great Programmer
- Python Prompt Strings
- Rule 30 - Generating Random Numbers with Cellular Automata
- Python's Missing Feature - Function Overloading
- Isolation Forest - Fast and Efficient Anomaly Detection
- Image Steganography
- How Python Handles Gigantic Integers
- Python Internals - I Made Addition Unpredictable