@hackproduct9: Most people learn these in the wrong order. ⚡ They start with Circuit Breaker because it sounds impressive. But the breaker only exists because retries made things worse — and retries only exist because of timeouts. 12 patterns, in the order they actually build on each other 👇 🟢 FAIL FAST 1. Timeout — cut it at 200ms, not 30s. A hung call still holds a thread. 2. Retry — backoff doubles: 1s, 2s, 4s. Jitter stops the thundering herd. 3. Circuit breaker — closed, open, half-open. One probe decides, not a timer. 🔵 PROTECT CAPACITY 4. Rate limiter — the 11th request gets a 429. Shed early, or shed everything. 5. Bulkhead — cart drowns, auth still serves. One pool per dependency. 6. Pub/Sub — one publish, three consumers. The publisher never waits. 🟡 CACHING & STATE 7. Cache aside — miss, fetch, then set. The app owns the cache, not the DB. 8. Write-through — no ack until the DB has it. Slower writes, never a stale read. 9. Event sourcing — replay the log, rebuild the state. Never overwrite, only append. 🟣 DISTRIBUTED, THE HARD PART 10. CQRS — the read side lags, on purpose. Scale reads without slowing writes. 11. Saga — step 3 fails, so undo 2 then 1. There is no distributed rollback. 12. Strangler fig — route by route, never a big bang. The old system dies last. 👀 The one nobody says: a retry without a breaker is an outage generator. Your dependency degrades, every caller retries, and the retries become the traffic that kills it. The rule: never add a retry without deciding when to stop retrying. Most engineers can name all 12 and can't say which two fight each other. That's the real interview question. 📸 Screenshot the grid. Save it before your next design round. Follow @hackproduct — we turn scary engineering concepts into things you can ship. ⚡ . . #systemdesign #softwareengineering #systemdesigninterview #backend #distributedsystems

HackProduct

Open In TikTok:

Region: US

Monday 10 August 2026 12:18:19 GMT