Databases at DoorDash | How DoorDash manages 1.9 PB of data & 1.2M QPS
summary
Runs CockroachDB at ~1.9 PB and 1.2M QPS across hundreds of clusters, single-region EC2 with replication factor 5 and ZFS-on-EBS. Uses Argo workflows for automated repaves and maintenance, cutting repave time from ~100 days to ~10 days, and an Auto Scaling + NLB front end for hitless node additions. Self-serve console and a metrics-exporter frontend expose SQL time series for query triage. Piloting Graviton m7g yields lower P99/P95 latency and ~15% lower list-cost versus m6i.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 0:32About the Core Infrastructure Team at DoorDash
- 1:08How CockroachDB is deployed at DoorDash
- 1:47Behind the numbers: CockroachDB at DoorDash
- 3:25Storage self-serve
- 4:36Operator Service + Automation
- 6:34Observability and optimizing for SQL efficiency
- 9:41The 1% Better Every Day principle
- 11:44Building for availability, efficiency, and performance
- 15:00Summary
related talks
31mFuture-proofing operational data at scale
15mScaling to support America’s #1 sportsbook
35mDoorDash’s Journey From StatsD To Prometheus With 10 Million Metrics/Second
32mPromoting a data driven culture in a world of microservices
28mHow Route enables always-on data for over 1 billion+ orders
43mFuture-proofing your applications with AWS databases
29mHow City Storage Systems designed its Data Infrastructure to minimize toil
48mBuilding mission-critical applications with CockroachDB