Introduction | Performance | Flexibility | Technology | Documentation | Download

The goal of Bullet cache is to offer the performance of 1,000,000 transactions per second on commodity hardware, measured as single-record get & put operations in 90% / 10% ratio, over dozens of clients. So far, it is quite near its goal, as the next graph shows.

This benchmark was performed on a single Xeon 3440 CPU running at 2.5 GHz. It is very likely that the goal has already been reached if a slightly faster CPU were used.

Bullet cache is heavily optimized for multiprocessor performance, going above and beyond standard techniques in avoiding multithreading performance bottlenecks, while at the same time using standard, verifiable algorithms.

Note to benchmarkers: Almost always, you want to specify the number of threads on the server by using the "-n" command line argument, not the "-t" argument!

Stay tuned for more high performance results!

Support This Project