Use sysbench or official benchmark kits to establish a reproducible OLTP baseline on your OceanBase cluster before hardware sign-off. Record TPS, latency percentiles, and per-node CPU at a fixed thread count, then re-run after topology or version changes.

Before signing a hardware purchase order, run a standardized benchmark that mirrors your transaction mix. This guide uses sysbench oltp_read_write as an OLTP stand-in; adjust table count and row size to match your schema width.

Cluster prep

  • Multi-node cluster with production-like zone layout.
  • OBProxy on separate hosts for realistic routing.
  • Dedicated benchmark tenant; document log stream count.

sysbench command (example)

sysbench oltp_read_write \
  --mysql-host=obproxy.bench.internal \
  --mysql-port=2883 \
  --mysql-user=root@bench \
  --tables=32 --table-size=5000000 \
  --threads=128 --time=600 \
  --report-interval=10 run

Replace host, port, and credentials with your environment.

Record

MetricPurpose
TPSRegression baseline
p95 latencySLA comparison
CPU per nodeHeadroom check

Scale threads until TPS plateaus. That plateau is an empirical ceiling for that topology, not a guarantee for other workloads.

Sources and further reading