This document discusses performance improvements and optimizations for MySQL NDB Cluster 8.0 based on TPC-H benchmark results. Key points include:
- Using NVMe disks, SQL and disk data performance is now almost on par with memory.
- Various configuration optimizations are shown to improve TPC-H latency for queries, including collocating MySQL servers and data nodes, optimized spinning, and increasing logical data memory partitions.
- NDB parallel query architecture can scale query performance up to 12 logical data memory partitions for most TPC-H queries, with future work aimed at further optimizations.