https://www.shaileshmishra.info/ Notes and references on the systems, languages and tools I work with - grouped so you can read a whole subject rather than a feed. en https://www.shaileshmishra.info/blog/system-design/23-case-studies.html https://www.shaileshmishra.info/blog/system-design/23-case-studies.html Four systems designed end to end - URL shortener, chat, news feed, key-value store - putting every part of the series to work at once. https://www.shaileshmishra.info/blog/system-design/22-deploys-rollback.html https://www.shaileshmishra.info/blog/system-design/22-deploys-rollback.html Blue-green, canary and rolling compared by blast radius, the mixed-version problem all three create, and why deploying is not releasing. https://www.shaileshmishra.info/blog/system-design/21-observability.html https://www.shaileshmishra.info/blog/system-design/21-observability.html Monitoring versus observability, the three pillars and the four golden signals, and why you alert on symptoms rather than causes. https://www.shaileshmishra.info/blog/system-design/20-distributed-locks.html https://www.shaileshmishra.info/blog/system-design/20-distributed-locks.html Pessimistic versus optimistic locking, what happens when a lock expires while the work is still running, and why fencing tokens are the fix. https://www.shaileshmishra.info/blog/system-design/19-consensus-leader-election.html https://www.shaileshmishra.info/blog/system-design/19-consensus-leader-election.html How a quorum makes split-brain impossible, what electing a leader actually involves, and when a system genuinely needs consensus. https://www.shaileshmishra.info/blog/system-design/18-rate-limiting.html https://www.shaileshmishra.info/blog/system-design/18-rate-limiting.html Token bucket versus leaky bucket, where the limit gets enforced, what the client sees when it trips, and what to key the counter on. https://www.shaileshmishra.info/blog/system-design/17-idempotency.html https://www.shaileshmishra.info/blog/system-design/17-idempotency.html Why a retried request bills the customer twice, what idempotent really promises, which HTTP methods honour it, and how keys cover the rest. https://www.shaileshmishra.info/blog/system-design/16-queues.html https://www.shaileshmishra.info/blog/system-design/16-queues.html What putting a queue in the middle actually buys you, queue versus pub/sub, the delivery guarantee you have to pick, and poison messages. https://www.shaileshmishra.info/blog/system-design/15-consistent-hashing.html https://www.shaileshmishra.info/blog/system-design/15-consistent-hashing.html Why naive mod-N hashing collapses on resize, how the ring fixes it, what virtual nodes add, and what it still doesn't solve. https://www.shaileshmishra.info/blog/system-design/14-sharding.html https://www.shaileshmishra.info/blog/system-design/14-sharding.html Partitioning strategies, the hot shard problem, what cross-shard queries cost, and how to pick a shard key you won't regret. https://www.shaileshmishra.info/blog/system-design/13-transactions-isolation.html https://www.shaileshmishra.info/blog/system-design/13-transactions-isolation.html ACID, the four classic anomalies, what each isolation level actually stops, and locking versus MVCC in real databases. https://www.shaileshmishra.info/blog/system-design/12-consistency-models.html https://www.shaileshmishra.info/blog/system-design/12-consistency-models.html What a reader is actually guaranteed to see - linearizability, eventual consistency, the middle ground, and what CAP forces during a partition. https://www.shaileshmishra.info/blog/system-design/11-replication.html https://www.shaileshmishra.info/blog/system-design/11-replication.html Synchronous versus asynchronous from the client's point of view, replication lag, failover when the leader dies, and the topologies between. https://www.shaileshmishra.info/blog/system-design/10-indexing.html https://www.shaileshmishra.info/blog/system-design/10-indexing.html Full scan versus index lookup, B-tree versus hash, the write cost every index charges, and why the planner sometimes skips yours. https://www.shaileshmishra.info/blog/system-design/09-what-a-write-actually-does.html https://www.shaileshmishra.info/blog/system-design/09-what-a-write-actually-does.html The write path step by step: why the WAL comes first, sync versus async acknowledgement, and what breaks mid-write. https://www.shaileshmishra.info/blog/system-design/08-cdn-edge.html https://www.shaileshmishra.info/blog/system-design/08-cdn-edge.html Why distance is the enemy, origin pull versus push, what belongs at the edge, and what being everywhere costs you. https://www.shaileshmishra.info/blog/system-design/07-caching.html https://www.shaileshmishra.info/blog/system-design/07-caching.html The cache hierarchy, read and write strategies, eviction policies, and the stampede that takes your origin down at the worst moment. https://www.shaileshmishra.info/blog/system-design/06-load-balancers.html https://www.shaileshmishra.info/blog/system-design/06-load-balancers.html Health checks, L4 versus L7, the algorithms behind each routing decision, and where a balancer sits in the request path. https://www.shaileshmishra.info/blog/system-design/05-stateless-vs-stateful.html https://www.shaileshmishra.info/blog/system-design/05-stateless-vs-stateful.html Where the state actually goes when you make a service stateless, and why sticky sessions are a trap rather than a solution. https://www.shaileshmishra.info/blog/system-design/04-vertical-vs-horizontal-scaling.html https://www.shaileshmishra.info/blog/system-design/04-vertical-vs-horizontal-scaling.html Bigger box or more boxes - the ceiling you hit with each, what horizontal scaling really costs, and the signals that decide it for you. https://www.shaileshmishra.info/blog/system-design/03-what-a-server-runs-out-of.html https://www.shaileshmishra.info/blog/system-design/03-what-a-server-runs-out-of.html CPU, memory, disk, network and file descriptors - the finite resources behind most outages, and how to tell which one gave out. https://www.shaileshmishra.info/blog/system-design/02-failure-taxonomy.html https://www.shaileshmishra.info/blog/system-design/02-failure-taxonomy.html How distributed systems actually break: partial versus total failure, why "slow" is harder than "dead", and the fault-tolerance toolkit for each. https://www.shaileshmishra.info/blog/system-design/01-latency-throughput-percentiles.html https://www.shaileshmishra.info/blog/system-design/01-latency-throughput-percentiles.html Why an average response time hides the experience most users actually get, and how p95 and p99 put it back on the table. https://www.shaileshmishra.info/blog/overview/system-design.html https://www.shaileshmishra.info/blog/overview/system-design.html What system design actually asks of you: turning an ambiguous product need into something reliable, scalable and explainable. Start here. https://www.shaileshmishra.info/blog/deep-dives/standalone--load-balancer-api-gateway-guide.html https://www.shaileshmishra.info/blog/deep-dives/standalone--load-balancer-api-gateway-guide.html Everything that sits between a client and your services: L4 versus L7, the routing algorithms, the gateway patterns, and which to reach for. https://www.shaileshmishra.info/blog/deep-dives/standalone-cache-complete-guide.html https://www.shaileshmishra.info/blog/deep-dives/standalone-cache-complete-guide.html The whole of caching in one place - where caches live, the read and write paths, TTL, eviction, and why invalidation is the hard part. https://www.shaileshmishra.info/blog/deep-dives/standalone-caching_investigation_board.html https://www.shaileshmishra.info/blog/deep-dives/standalone-caching_investigation_board.html One overworked database worked like a case file: why the problem exists at all, how caching closes it, and where the trail goes cold. https://www.shaileshmishra.info/blog/deep-dives/system-design-concepts-and-patterns.html https://www.shaileshmishra.info/blog/deep-dives/system-design-concepts-and-patterns.html The whole vocabulary in one place: non-functional requirements, estimation, scaling, caching, SQL versus NoSQL, and the patterns built on them. https://www.shaileshmishra.info/blog/deep-dives/messaging-async-systems-guide.html https://www.shaileshmishra.info/blog/deep-dives/messaging-async-systems-guide.html Queues, pub/sub and event-driven architecture end to end - RabbitMQ against Kafka on their real differences, and which one a problem wants. https://www.shaileshmishra.info/blog/deep-dives/security-in-system-design-guide.html https://www.shaileshmishra.info/blog/deep-dives/security-in-system-design-guide.html Authentication versus authorisation, what a JWT does and does not prove, encryption, rate limiting, and the design habits that prevent the rest. https://www.shaileshmishra.info/blog/deep-dives/dns-complete-guide.html https://www.shaileshmishra.info/blog/deep-dives/dns-complete-guide.html How a name becomes a connection: the resolution chain, record types, why caching keeps the root idle, and GeoDNS as a load-balancing tool. https://www.shaileshmishra.info/blog/deep-dives/database_investigation_board.html https://www.shaileshmishra.info/blog/deep-dives/database_investigation_board.html One machine that ran out of room, worked like a case file - where the data goes next, and what each escape route costs you. https://www.shaileshmishra.info/blog/deep-dives/networking_investigation_board.html https://www.shaileshmishra.info/blog/deep-dives/networking_investigation_board.html Two machines a thousand miles apart and one lost message: the layers side by side, and who really stands where among proxies and gateways. https://www.shaileshmishra.info/blog/ai/ai-systems-study-guide.html https://www.shaileshmishra.info/blog/ai/ai-systems-study-guide.html Seventeen topics built as one project - embeddings, chunking, context engineering, agents, guardrails, evals - each with runnable code and a cost lens. https://www.shaileshmishra.info/blog/ai/the-agentic-ai-roadmap.html https://www.shaileshmishra.info/blog/ai/the-agentic-ai-roadmap.html What lets a model use your own data: how embeddings encode meaning, why chunking is the real quality lever, and where vector search fits in RAG.