Research on Performance Testing Techniques for Relational Database
The current TPC-E benchmark testing system suffers from continuous aggregation of the same type of transactions and test truncation,which can lead to poor test transaction cases and inaccurate test results.Therefore,a transaction distribution control method based on transaction mixed queue and delayed end strategy is proposed to en-sure that the test transaction use case ratio is as close to the benchmark as possible from the source.The transaction mixing queue implements the overall control of the over the mixing ratio of test transaction cases.The delayed end strategy eliminates the bumping of the transaction mix ratio caused by test tail-breaking.In addition,a dynamic thread pool model is proposed to reduce resource consumption.This model can effectively improve system response when there are too many requests and effectively utilize system resources when there are too few requests.Finally,the queue caching mechanism is introduced into the test system.Caching the request to send highly concurrent things in the queue alleviates the problem of instantaneous resource preemption caused by the creation of a large number of threads.Experiments verify the effectiveness,stability and suitability of the test system.
Transaction distribution controlDynamic thread poolQueue cacheHigh concurrency
NETL
NSTL
万方数据
