We ran iiBench on the same hardware (Sun x4150, 8 cores @ 3.16GHz, 16GB memory, 6 SAS disk HW RAID 0) using Tokutek’s storage engine for MySQL. The performance looks like
Our engine consumed 20.2 hours elapsed time, and inserted the last 10M rows at a rate of 11,220 rows/second. In contrast, InnoDB finished the test in 207.5 hours, inserting the last 10M rows at a rate of 876 rows/second. For our engine, the my.cnf file has no special parameters defined – the test was run with default parameters. By default, Tokutek uses 50% of the physical memory for its buffer pool. That works well for a broad range of physical memory sizes, eliminating the need for tuning parameters in many cases. iiBench sets a single Tokutek-specific variable, tokudb_commit_sync=off, in the client session. Setting this variable defers synchronization of the log files, similar to setting innodb_flush_log_at_trx_commit=2 for InnoDB.
Take a look at our early release program if you’re interested in evaluating the technology.