System Design Simulator

Turn up the traffic. Find the bottleneck. Break things. · made by @jeremybernier

Incoming0req/s

Served0req/s

Errors0%0 req/s

Bandwidth0through load balancer

Est. cost$0per month

Latency p500p99 —

SLO100%error budget

System

Workload assumptions

What an average request looks like. These describe your product, not any technology; hover a row for what it changes.

Requests per second Pattern

Writes

Average asset size CDN cache TTL

Web servers

Autoscale web servers

Min servers

Max servers

Target CPU

Autoscale workers on queue depth

Workers

Database read replicas

Database shards

Cache nodes

Auto-restart crashed nodes

Client timeout Client retries Retries per request, at most SLO: requests that must succeed SLO: p99 latency under

Dashboard queries per second Iceberg compaction

Learn