RocolSys: An Automatic Row-Column Data Storage System for HTAP
摘要
Hybrid Transactional and Analytical Processing (HTAP) brings great challenges to data storage systems. However, traditional approaches have struggled to dynamically adapt storage structures to data and complex workloads. Fortunately, machine learning can provide new methods to guide decisions on data storage structure selection. Motivated by this, we develop RocolSys, an automatic hybrid data storage system for HTAP that can predict workloads and select storage structure automatically based on machine learning. RocolSys predicts workloads accurately and selects storage structures efficiently. It also provides the user-friendly interface that allows users to connect to their own databases. We demonstrate the efficiency of RocolSys on public benchmarks.