Design And Implementation Of Automatic Integration Testing Platform For Distributed Environment

With the enlargement of the software, the software testing takes a moreimportant place in the link of software quality assurance, baidu s recommendationand personalization related products product composition is complex, testing workiterates frequently. Testers for regression testing, developers for product research,product staff for product evaluation of the effectiveness, usuall y involve a series ofnecessary, tedious and error-prone work. How to improve different roles of staff swork efficiency is a main research for this subject, so baidu s recommendation andpersonalization department decide to develop a set of automated integration testingplatform based on characteristics of their own product to solve the problem above.Considering that the integration testing platform need to run in baidu s clusterenvironment, and the products tested are also executed in baidu s distributedenvironment. The subject mainly involves task management and distribution, settingup test environment, execution of product flow, execution of automated test cases:Considering server s performance, robustness and implementation cost in theintegrated testing platform s design and implementation. The integrated testingplatform uses distributed remote procedure processing framework which is calledGearman to achieve the submitted tasks management and distribution. By usingmysql database to achieve uncompleted and pending tasks persistence work inorder to recover the unfinished task to from the machine fault.The integrated testing platform have implemented pulling environment fromthe online machine, modifying from online configuration to offline configuration,and doing the localization work for the specific module. The script of building theenvironment shoud be accumulated in the development and testing process, andunified interfaces are provided to facilitate integration.Based on the built offline environment, The integrated testing platform callbaidu s recommendation and personalization department s self developed flowengine. Through the flow engine s interface to get the last executed task flow s lastexecuteds algorithm package s execution status, in order order to decide the furtherprogram flow.After taking into account the function extensibility of testing framework andthe tested products characteristics in the integrated testing platform s design andimplementation. the testing platform adopts python automated testing frameworkcalled nose.Based on the above research content, against baidu recommendations andpersonalization products, the integrated test platform provides a user-friendly front-end web access, integrated interface for product developers. The distributedarchitecture provides the flexibility to respond to future recommendation andpersonalization product s architectural change.