Research And Implementation Of Benchmark For Massive Database

With rapid development of information society, a huge amount of data were produced and stored in database. The size of problem leads to the emergence of massive database. With the volume of database storage scaling magnitude of orders, the first thing building massive database is how to assure the performance of system to meet customer's performance requirements.Application-oriented benchmark can mimic work load of real world to evaluation the system's current performance. With appropriate designment of test case and performance analysis techniques, we can predict the future performance of system. Through repeated executions of thousands of virtual users'test, we can locate the performance bottleneck, then tune performance. So, building application-oriented benchmark can effectively resolve the performance problem of building massive database.This paper first introduces the-state-of-the-art massive database and the industry-standard series of TPC benchmarks. Then, based on the characteristics of massive data process system, we propose a massive database oriented benchmark architecture. Then, we research the key technologies of how to implement the benchmark. Then, we research how to tune system performance with the benchmark. After that, how to develop benchmark test suites in the early stage of system was researched. At last, we benchmarked a specific massive database based on StarTpMonitor to help developer and decision-maker to evaluate system performance.