Problems / Caching, hard

Everyone reads something different

Popularity is nearly flat (Zipf 0.4) over 10 million keys, so a small cache barely helps.

Free to open. Runs in your browser, no sign-up.

Your goals

Success ≥ 99%, p99 ≤ 300 ms.

2,500 requests a second for 30 seconds: 90% reads, 10% writes.

What each component does in this design, and what calls it. The simulator mimics each one from these settings.

Component	Its job here	Settings
Load balancerLoad balancer	Traffic enters here.Calls API.Layer 4 balancers route by IP and port; layer 7 balancers read HTTP and can route by path or header.	–
APIApp service	Load balancer forwards to it.Calls Redis cache (checked first), Postgres.Little's law: throughput ≈ concurrency ÷ latency. 200 slots at 50 ms serve about 4,000 req/s.	3 replicas × 200 slots · timeout 1.0 s
Redis cacheCache	API checks it first for reads.An in-memory cache like Redis usually answers in well under a millisecond inside one region.	2 GB memory
PostgresSQL database	API reads it when Redis cache misses. API calls it for writes.Postgres runs one process per connection, which is why poolers like PgBouncer sit in front of busy databases.	2 vCPU · 40 connections

Try it first. Open one hint at a time if you get stuck.

Hint 1

With flat popularity the hit rate ≈ the share of keys that fit.

Hint 2

Either fit much more, or scale the database.

One way to solve it

Give the cache enough memory for most keys, or add read replicas.