Scaling Postgres With Some Help from Redis
summary
Combining Postgres for durable storage with Redis for high‑throughput in‑memory work proves concrete: Redis benchmarks at ~300k pipelined ops/sec on one core, a search index goes from idea to production in weeks and returns results in ~15 ms, an ad matcher achieves ~20 ms internal / 50 ms end‑to‑end, and a streaming analytics pipeline handles ~75M events per hour (about 20k writes/sec) and pulls 12.5M items in under 5 minutes while syncing aggregates back to Postgres.
AI-written summary · about
chapters (19)
- 0:00Start
- 1:38Who Am I
- 10:26What Does Postgres Good at
- 11:31What Are some Limitations of Postgres
- 17:51Server-Side Scripting with Lua
- 19:35Event Analysis
- 24:03Redis Is a Search Engine
- 28:40Storage for a Twitter Analytics Engine
- 29:18Why We Ended Up Using Redis
- 32:42Cookie Storage
- 34:26What Kind of Things Does Redis Do Really Well
- 36:27Data Durability
- 38:14What Is Redis
- 39:11Redis Is a Shared Memory Api Server
- 40:52When Should I Use Redis
- 42:50Emulate Tables and Rows
- 45:47Sorted Sets
- 47:55How Different Is Redis from no Sequel
- 48:31Alchemy Db
