BestConfig

BestConfig is an automatic configuration tuning system. It can automatically find, within a given resource limit, a configuration setting that can optimize the performance of a deployed system under a specific application workload.

BestConfig features a highly flexible and extensible architecture, the scalable sampling method DDS (Divide-and-Diverge-Sampling) and the scalable performance optimization algorithm RBS (Recursive-Bound-and-Search).

BestConfig has successfully been applied to tune the configuration settings of six widely used systems, with the obvious performance improvements observed after tuning. Tuning the Tomcat system on virtual machines in the Huawei cloud, BestConfig has actually made it possible to improve the performance of a fully loaded system by simply adjusting its configuration settings. These results highlight the importance of an automatic configuration tuning system for tapping the performance potential of systems.

BestConfig is expected to benefit users in the following five aspects:

  1. Improving system performance: as much as 11 times performance gain
  2. Improving system utilization: eliminating 1 virtual machine need from every 26
  3. Saving labor costs: work for 5 employees half a year tuned into that for machines in two days
  4. Enabling fairer benchmarking and comparison of systems
  5. Identifying system bottlenecks

Publications

  • Yuqing ZHU, Jianxun Liu. ClassyTune: A Performance Auto-Tuner for Systems in the Cloud. IEEE Transactions on Cloud Computing. doi: 10.1109/TCC.2019.2936567 (pdf).
  • Yuqing ZHU, Jianxun Liu, Mengying Guo, Yungang Bao, Wenlong Ma, Zhuoyue Liu, Kunpeng Song, Yingchun Yang. BestConfig: Tapping the Performance Potential of Systems via Automatic Configuration Tuning. Proceedings of the ACM Symposium on Cloud Computing 2017 (SoCC’17) (pdf, slides).
  • Yuqing ZHU, Jianxun Liu, Mengying Guo, Yungang Bao. ACTS in Need: Automatic Configuration Tuning with Scalability. Proceedings of the 8th ACM SIGOPS Asia-Pacific Workshop on Systems (APSys’17) (pdf).

Source Code

As an open-source package, BestConfig is available for developers to use and extend in order to effectively tune cloud systems.

Talks

BestConfig is presented at SoCC'17 (slides), APSys'17 and HotDC (slides).