Research On Performance Test Of DBMS With TPC-W Benchmark

With the wide application of Ecommerce, the research on high performance DBMS used in Ecommerce environment becomes a hot field, where the performance test of the Database Management System (DBMS) plays a key role. Database benchmarks defined by TPC have been adopted widely. Based on the TPC Benchmark?W (TPC-W), the theory, the method and the implementation of performance test among different DBMSs such as DM4 and Oracle- 9i are systematically proposed. Applying the TPC-W benchmark targeted at the Ecommerce environment to evaluating the performance of DM4 and comparing its results with those of other DBMSs, we can make improvements to DM4.After analyzing the characteristics of web load and the theoretic foundation, a DBMS performance testing model targeted at the Ecommerce environment is presented, it includes the primary metrics such as throughput, response time, think time and so on, the testing method and the testing system architecture. The DBMS performance testing system of TPC-W——DM-TPCW is designed and implemented based on it. According to TPC-W, the implementation of the remote browser emulator, the logical design of database in system under test, the web transaction model are narrated. The performance test system with TPC-W composes five modules: data generation and loading, the database transaction ACID properties test, the web page consistency test, the performance test and results statistics. Furthermore, the database transaction ACID properties, the requirements and the testing method of web page consistency are presented.DM-TPCW makes following improvements. First, the improved data generator DM-DBLOADER can customize data generation methods, generate and import data quickly. Then it can customize transaction types and schedule plan, so DM-TPCW can implement test for different working modes easily.To compare the performance of DM4 with that of other DBMSs, we use DM-TPCW to test DM4, Oracle9i and Microsoft SQL Server 2000 separately, analyze the results, and give suggestions for the improvement to DM4's performance, such as the query schema based on the optimization technology, the waiting time of the transaction restarting.