Benchmark Your Cloud Native Database
summary
Compares four storage types on a six‑blade bare‑metal cluster (20 cores, 128 GB RAM, two 200 GB SSDs): bare metal host, container with local PVs, and cloud‑native distributed storage (Rook/Ceph). Runs microbenchmarks using sysbench, pgbench, and CockroachDB TPCC; tests fit‑in‑memory workloads so file‑sync latency dominates. Findings: local volumes match bare metal within ~10–15% except when extra network hops inflate latency; Rook gives similar throughput for many reads but roughly 2× latency and about 50% throughput on random write/transaction tests. Advises multiple long runs and tuning for your hardware.
AI-written summary · about
chapters (12)
- 0:00Benchmark Your Cloud
- 3:07why performance?
- 4:41benchmarking = comparing
- 5:23types of storage
- 8:46DB (micro) benchmarks
- 10:16postgres pg_bench
- 10:40cdb workloads
- 12:58why bare metal?
- 16:41host filesystem
- 19:56local volumes test
- 22:43rook storage
- 26:51conclusions
related talks
32mBenchmark Your Cloud Native Database
39mBenchmarking Cloud Native Storage
56mBenchmark Your Cloud-Native Database
05mIgnite - How Fast Was My Kubernetes?
33mBenchmarking Cloud Native Databases Performance on Kubernetes
17mComparing DBs with Golang in the cloud
47mThe Road Towards an Open Database Performance Ranking
36mBest Practices in PostgreSQL Tuning: Navigating Key Performance Bottlenecks in the Cloud