Optimizing Concurrent Query Execution on Modern Multisocket Multicore Platform

The upcoming generation of computer hardware brought several new challenges for underling software. The number of cores on a chip grows exponentially, allowing for an ever-increasing number of processes to execute in parallel. Therefore, how to efficiently utilize all of the concurrent processing ability provided by multi-core platform becomes critical for achieving good system performances. We propose a CARIC-DA framework which optimizes the performance of DBMS on multicore platforms. Our middleware can improve the performance of the private cache levels by dispatching concurrent database queries to run on different processor cores according to the data needs of each query. Extensive experiments from microbenchmark to TPC-C benchmark have confirmed the effectiveness of CARIC-DA in improving the cache performance and providing higher throughput for DBMSs on modern AMD multicore platform. However, the multicore platforms are becoming more and more diverse and complex in recent years. Different manufactures are making different efforts to boost their performance by adopting different techniques. Therefore, it is important for software to be well turned according to the specific hardware features in order to fully take advantage of the target multicore platforms. In this paper, we analyzed the efficiency of our CARIC-DA framework on the modern multicore Intel platform. We compared the performance of different deployment strategies where we vary the mapping relationship between database processes and processor cores from Hyper-Threading enable to disable configurations. We evaluated the impact of CARIC-DA for different cache levels and for different query types on the Intel multicore platform.