Adaptive Virtual Partitioning for OLAP Query Processing in a Database Cluster

摘要: OLAP queries are typically heavy-weight and ad-hoc thus requiring high storage capacity processing power. In this paper, we address problem using a database cluster which see as cost-effective alternative to tightly-coupled multiprocessor. We propose solution efficient query simple data parallel technique called adaptive virtual partitioning dynamically tunes partition sizes, without any knowledge about the DBMS. To validate our solution, implemented Java prototype on 32 node system ran experiments with typical of TPC-H benchmark. The results show that yields linear, sometimes super-linear, speedup. many cases, it outperforms traditional by factors superior 10.