Research On Key Technology Of TPC-E Benchmark Test System

In recent years,with the advancement of technology,the quality requirements of software products have become more stringent.Especially under the situation of epidemic prevention and control,all kinds of businesses are turning to online completion,making system performance face greater challenges.As the heart of the system,the database is an indispensable part of any enterprise or individual.Faced with many database types,choosing a suitable database has become a problem.The iteration cycle of database product updates is getting shorter and shorter.The pace of innovation is getting faster and faster with the pace of the times.It is even more difficult for users to make purchasing decisions.difficult.Even if the database can meet the existing requirements,when applied to high-load and large-scale scenarios,users still have doubts,and the performance of the database needs to be verified.At present,compared with mainstream foreign databases,there are still some gaps in the design of domestic databases.The core technology is restricted by others.Database performance testing can explore system bottlenecks,locate problems and analyze defects,and help independent innovation of core technologies.Choosing a recognized authoritative testing method,testing the database,and evaluating the database performance scientifically and fairly is an important mission in the field of database performance testing.Based on this,this thesis selects the international TPC-E benchmark as the test method guide,and deeply interprets the TPC-E benchmark from the three elements of test cases,test execution,and test results,forming constraints around business models,data models,execution models,and evaluation models.Restricted requirements,combined with the requirements of the TPC-E benchmark test system and the tool components provided by the official,proposed two key technologies to optimize the realization of the TPC-E benchmark test system,namely the transaction distribution control method and the result data processing method.The transaction distribution control method solves the problem of data aggregation and distortion through the combination of queues.The delayed end strategy efficiently ensures that the transaction mixing ratio meets the benchmark requirements,and prevents the impact of test tailing from the source to the greatest extent.The test result data processing method uses a fine-grained sliding window method to select the best results from the test results that meet the requirements for reporting,which solves the problem of system performance interpretation errors caused by improper test result data processing,and more accurately evaluates database performance.Finally,the above two methods are applied to the TPC-E benchmark test system to verify the effectiveness of the method and achieve good application effects.