Implementing TaaS-based stress testing by MapReduce computing model

Abstract

In this paper we propose to employ the MapReduce computing model to implement a Testing as a Service (TaaS) for stress testing. We focus on stress testing for heavy-weight network transactions. The computation power of MapReduce computing system is used to simulate concurrent network transactions issued by a lot of users. The user first describes the testing scenario which he wants to be simulate in a testing script. The TaaS platform analyzes the testing script and then automatically distributes required testing data including details of transactions and files into a MapReduce computing system. We propose three different schemes to distribute testing data and measure their performances. We compare them with the popular stress testing tool JMeter and find out that our scheme can always have the tested system deliver higher error rate during the stress testing.