Essential Tips for Optimizing Chinese Domestic Databases: What Oracle DBAs Should Know

Introduction

As domestic databases become increasingly common in China, DBAs who have worked primarily with Oracle need to understand how optimization practices differ. Benchmark numbers such as TPMC are often advertised, but they can be misleading without proper context.

Key Precautions

Do not blindly apply the parameter adjustments used in vendor‑provided benchmarks. Many settings that improve benchmark scores are unsuitable for production and may harm stability.

Ignore traditional Oracle buffer‑hit‑rate metrics. For most domestic databases, which are derived from PostgreSQL or MySQL, buffer‑hit ratio does not accurately reflect real performance. Focus instead on OS‑level buffers, dirty‑page flush policies, virtual memory allocation, and NUMA tuning.

Prioritize SQL execution efficiency. The cost‑based optimizer (CBO) in many domestic products lags behind Oracle. Build a performance baseline for critical SQL statements, regularly compare execution plans, and address plan regressions promptly.

Address high‑concurrency challenges. Issues such as hot‑block conflicts and spin‑lock contention are common. Adjust data‑block fill‑rate parameters, consider hash partitioning for large tables, and tune spin‑lock settings using tools like perf to reduce wait times.

Practical Observations

Increasing the database buffer size often yields less noticeable gains than tuning OS buffers and I/O latency. Stable, low‑latency physical I/O on the operating system is crucial for overall concurrency and query performance.

Conclusion

Optimization of domestic databases is still an exploratory field, and vendor‑provided guidelines are typically rudimentary. As adoption grows, best‑practice knowledge will evolve, but DBAs must continue to experiment, share findings, and encourage vendors to focus on real‑world performance tuning rather than solely on benchmark bragging.