High Level Design
System design, architecture patterns, scalability, and distributed systems. How to design systems that handle millions of users.
SSL / TLS Explained
HTTP vs HTTPS, TLS handshake, TLS 1.3, certificates & CAs, symmetric vs asymmetric encryption, AES-256, cipher suites, mTLS, attacks & mitigations, and best practices.
JWT & Auth
How JSON Web Tokens actually work — structure, signing and verification, auth flows, access and refresh tokens, and the security pitfalls, compiled into one unit.
Netflix Hollow — Deep Dive
How Netflix Hollow replicates a read-only, in-memory dataset from one producer to many consumers — snapshots, deltas, reverse deltas, ordinal-based columnar storage, producer/consumer lifecycle, and failure modes.
Dynamo: Amazon's Highly Available Key-value Store
Amazon's 2007 Dynamo paper — a decentralized, highly available key-value store: consistent hashing, vector clocks, sloppy quorums, hinted handoff, Merkle trees, and gossip-based membership.
Cluster Architecture — Building Up the Ring
Build a Cassandra cluster from one node up — naive partitioning, consistent hashing, virtual nodes (vnodes), replication, and rack-aware placement, derived from first principles.
Consistency
Tunable consistency in Cassandra — why different operations need different correctness, quorums, R + W > N, read repair, and consistency levels.
Cassandra Storage Engine Internals
Why Cassandra uses LSM trees instead of B-trees — memtables, SSTables, the commit log, compaction, bloom filters, and the read/write paths.
How Discord Stores Billions of Messages
How Discord scaled message storage to billions on Cassandra — data modeling, partitioning by bucket, and the operational lessons learned.
How Discord Stores Trillions of Messages
How Discord moved from Cassandra to ScyllaDB to store trillions of messages — GC pain, hot partitions, and a shard-per-core C++ rewrite.
Multi-Paxos Deep Dive
How Multi-Paxos elects a stable leader to turn single-value Paxos into an efficient, fault-tolerant log-replication engine for distributed databases.
Nginx Deep Dive
How NGINX serves hundreds of thousands of concurrent connections on modest hardware with a small pool of event-driven worker processes.
How Amazon S3 Works — Durability & Scale
How Amazon S3 reaches 11 nines of durability and serves over 1 PB/s by erasure-coding objects into shards spread across millions of cheap drives.
The October 2025 DynamoDB us-east-1 Outage
How a latent race condition in DynamoDB's automated DNS management emptied the us-east-1 endpoint's DNS record and cascaded into a ~14.5-hour AWS outage.
Dynamo & Bigtable Inheritance
How Cassandra inherits its cluster mechanics from Amazon Dynamo and its on-disk storage engine from Google Bigtable, replacing vector clocks with last-write-wins.
Cassandra Data Modeling & Anti-Patterns
Query-first data modeling in Cassandra — hot partitions, partition-size limits, common anti-patterns to avoid, and when not to use Cassandra.