#architecture
33 posts
-
1 min read
#DistributedDilemmas: all 24 days in one place
X In February 2025 I posted one distributed systems dilemma a day on X. I made it to day 24. Here they all are, with the diagrams.
-
1 min read
Day 24 of #DistributedDilemmas: Distributed sessions are a persistent headache X Distributed sessions are a persistent headache. Sticky sessions don't scale, but fully distributed sessions add complexity. Consider JWT or externalized…
-
1 min read
Day 23 of #DistributedDilemmas: Distributed caching layers often become consistency nightmares X Distributed caching layers often become consistency nightmares. As cache invalidation schemes attempt to maintain coherence, real-world implementations…
-
1 min read
Day 22 of #DistributedDilemmas: Distributed background jobs need careful orchestration X Distributed background jobs need careful orchestration. While stateless architectures simplify scaling, they risk duplicate processing like Slack alerts…
-
1 min read
Day 21 of #DistributedDilemmas: Event sourcing isn't a silver bullet for distribution X Event sourcing isn't a silver bullet for distribution. Replay and snapshots become bottlenecks at scale. Design for efficient event storage and retrieval…
-
1 min read
Day 20 of #DistributedDilemmas: Distributed rate limiting is harder than it looks X Distributed rate limiting is harder than it looks. Local rate limits don't prevent global abuse. Global protection requires coordination mechanisms like…
-
1 min read
Day 19 of #DistributedDilemmas: Database migrations in distributed systems require meticulous planning due to their inherent complexity X Database migrations in distributed systems require meticulous planning due to their inherent complexity. Schema changes become distributed coordination…
-
1 min read
Day 18 of #DistributedDilemmas: Microservice authentication is a distributed minefield X Microservice authentication is a distributed minefield. While JWTs excel at decentralized authentication, they introduce authorization complexity due to…
-
1 min read
Day 17 of #DistributedDilemmas: Distributed cron jobs are a coordination nightmare X Distributed cron jobs are a coordination nightmare. Time drift and job duplication plague naive implementations. Consider distributed schedulers or…
-
1 min read
Day 16 of #DistributedDilemmas: Serverless doesn't mean worry-free X Serverless doesn't mean worry-free. Cold starts and hidden state create new classes of distributed problems. While offering auto-scaling and cost…
-
1 min read
Day 15 of #DistributedDilemmas: Distributed logging creates more noise than signal X Distributed logging creates more noise than signal. Modern distributed systems face two critical logging challenges: 1. Data overload from scattered logs…
-
1 min read
Day 14 of #DistributedDilemmas: Service discovery often becomes a SPOT (single point of failure) X Service discovery often becomes a SPOT (single point of failure). Your registry being unavailable shouldn't prevent finding the registry. Design for…
-
1 min read
Day 13 of #DistributedDilemmas: Configuration sprawl is the silent killer of distributed systems X Configuration sprawl is the silent killer of distributed systems. Environment variables become unmanageable at scale. Treat config as code, not as an…
-
1 min read
Day 12 of #DistributedDilemmas: API versioning perpetuates distributed mistakes X API versioning perpetuates distributed mistakes. v2 endpoints often carry v1's technical debt. Evolve APIs additively to avoid version hell.
-
1 min read
Day 11 of #DistributedDilemmas: Queue-based architectures hide scaling issues X Queue-based architectures hide scaling issues. Kafka topics with long retention become both a blessing and a curse. Backpressure beats unbounded queues!…
-
1 min read
Day 10 of #DistributedDilemmas: Observability becomes obscurity at scale X Observability becomes obscurity at scale. More metrics often mean less visibility. Focus on the critical few signals that truly indicate system health!…
-
1 min read
Day 9 of #DistributedDilemmas: Distributed transactions are a myth at scale X Distributed transactions are a myth at scale. Two-phase commit falls apart under network partitions. Design for compensation, not for distributed…
-
1 min read
Day 8 of #DistributedDilemmas: Eventual consistency is consistently misunderstood X Eventual consistency is consistently misunderstood. CRDTs work better in whitepapers than in production. Real-world systems need a spectrum of consistency…
-
1 min read
Day 7 of #DistributedDilemmas: Network partitions heal dangerously X Network partitions heal dangerously. Merging divergent states introduces subtle data corruption. CAP theorem isn't a choice, it's a law of nature in…
-
1 min read
Day 6 of #DistributedDilemmas: Leader election creates false security X Leader election creates false security. Raft/Paxos implementations often have higher MTTR than the failures they aim to solve. Simplicity frequently beats…
-
1 min read
Day 5 of #DistributedDilemmas: Sharding decisions made today become tomorrow's scaling bottlenecks X Sharding decisions made today become tomorrow's scaling bottlenecks. NoSQL's partitioning promises often cost more than vertical scaling. Distribution ≠…
-
1 min read
Day 4 of #DistributedDilemmas: Idempotency is a lie we tell ourselves X Idempotency is a lie we tell ourselves. Nested partial failures expose retry logic flaws. True exactly-once delivery remains elusive in distributed…
-
1 min read
Day 3 of #DistributedDilemmas: Caching accelerates inconsistency X Caching accelerates inconsistency. Cache invalidation storms often outweigh hit rates. The fastest cache is sometimes no cache at all. Exhibit A: DoorDash…
-
1 min read
🚀 Launching today: 28 Days of #DistributedDilemmas
Medium Follow me on Twitter/X this February for one battle-tested insight a day on the hidden complexities behind outages, tech debt and architectural nightmares.
-
1 min read
Day 2 of #DistributedDilemmas: Microservices don't eliminate complexity, they redistribute it X Microservices don't eliminate complexity, they redistribute it. Your monolith's function calls become network calls. Pick which failures you want to debug…
-
1 min read
Day 1 of #DistributedDilemmas: Consistency guarantees become availability nightmares X Consistency guarantees become availability nightmares. Every distributed lock is a future bottleneck. Systems that embrace controlled inconsistency often…
-
8 min read
Cell-based architecture, explained: the public version of my Innovation Day talk X My Innovation Day deck was internal, so here's the public version: what a cell is, scaling up and across, blast radius, and what cells cost you.
-
11 min read
The Redis Paradox: my DelhiFOSS 2.0 talk at IIT Delhi, written up X The written version of my DelhiFOSS 2.0 talk on Redis beyond caching: a short history of caches, live driver matching with GEO commands, and Pub/Sub.
-
6 min read
Speccing a Shopify app for a friend: Etsy-to-Shopify inventory sync
Obsidian A friend needed a Shopify app that keeps stock in sync with Etsy. Before any code, I wrote him a spec to hand to a developer: scope, diagrams and a 70-hour initial estimate.
-
4 min read
Our Journey with Multi-region Cloud Architecture and Data Residency
LinkedIn How Atlassian built its multi-region cloud architecture for data residency, covering region creation, service hydration, tenant migration and self-serve moves.
-
9 min read
Elevating Trust: Atlassian's Journey to HIPAA Eligibility Expansion! LinkedIn How Atlassian expanded HIPAA eligibility by replacing a manual BAA process with a self-serve flow and a new service integrating IronClad with Admin Hub.
-
9 min read
Two system design problems from my 2022 interviews, worked with real numbers Obsidian A photo pipeline for a fleet of 360-degree camera cars and an A/B split-testing service: requirements, capacity math, data model and the parts I'd tighten.
-
16 min read
Interview prep, senior edition: a system design checklist, one backtracking template and what loops grade on Obsidian My 2022 prep notes for senior and staff interviews: a system design checklist, one Python DFS for seven backtracking problems, rubrics and a STAR story bank.
