Identifying Workloads in Mixed Applications
Jeong Seok Oh, Hyo Jung Bang, Yong Do Cho · 2013
Database administrators should be aware of workload characteristics for managing database systems. Workload characteristics can be different depending on database application. In particular, identifying workloads in mixed database applications might be quite difficult. Therefore, a method is necessary for identifying workloads in the mixed database application. This paper aims to identify workloads in the mixed database application using data mining technologies. To construct the mixed database application, we use the TPC-C and TPC-W benchmark. We discriminate between train workloads (workloads using the TPC-C or TPC-W benchmarks) and test workloads (mixed workloads of both benchmarks), and modify the algorithm of k-NN (Nearest Neighbor) classifier in order to satisfy our objectives. The modified k-NN algorithm measures how close the test workloads are to the train workloads. The modified k-NN algorithm is better than others for identifying workloads because its results are lower than others in the oscillation depending on the k parameter and the error rate. This research contributes towards considering flexible tuning methods using workload identification information