Avoiding Overload in Distributed Systems
summary
Shows concrete ways to keep goodput flat under overload: load shedding using CPU/latency/connection heuristics and an SDK circuit breaker to stop harmful retries; per-request isolation via AWS Lambda execution environments so each invocation gets fixed CPU/memory and latency stays flat; queue fixes using priority queues and moving messages (cheaper than full processing) to simulate LIFO behavior; and multi-tenant isolation with shuffle sharding (example: 8 queues with k=2 gives ~3.6% full overlap; 100 queues with k=5 drops overlap to negligible), plus autoscaling and backpressure.
AI-written summary · about
chapters (6)
- 0:00Welcome
- 1:26Avoiding overload in distributed systems
- 8:10Load shedding techniques
- 13:47Queue systems and backlog
- 20:45Shuffle sharding
- 23:56Summary
related talks
52mWhy today, more then ever you need Dependency Isolation in Distributed Systems
34mDistributed Systems - What Can Go Wrong and How to Get It Right
02h 42m!!! Do not publish or share !!!!
40mEverything Will Flow
01h 01mLittle things and big things: lessons for building large systems.
13mExplaining Distributed Systems Like I'm 5
21mHandling Traffic Spikes Through Automated Prioritized Load Shedding
46mPracticalities of Productionizing Distributed Systems