Little's law based validation framework for load testing

doi:10.1016/j.infsof.2018.11.007

Little's law based validation framework for load testing

2019-04-01

4

PRE

AI

DOI：10.1016/j.infsof.2018.11.007

原文链接

原文求助

摘要

En 中文

Context: Performance is a key quality consideration for large-scale software systems which supports thousands of concurrent users. Load testing is an integral part of the development lifecycle and is used to address performance issues before deploying the system in production. But, how do we validate the output reported by load testing tools? Little's Law is useful for validating the accuracy of load testing output. Though IT industry is flooded with various enterprise and open source tools for load testing, but they do not offer support for validation of its result using Little's Law. Manual validation is time intensive and infeasible with increase in the complexity of testing scripts. Objective: In this paper we provide a Little's law-based validation framework which will enable the researchers and industry practitioners to validate load testing results. The implementation of the framework is also demonstrated. Method: To understand the constructs commonly used in load test scripts, we analysed scripts of two open source benchmark applications and eight large-scale software systems used in industry. We found that transactions are arranged using control flow patterns like sequential, loop and conditional. Based on the analysis, we devised the framework. Results: The efficacy of the proposed framework is successfully evaluated on two systems - an open source Dell DVD Store benchmarking application and a real-world large scale system used in industry. The framework is independent of load testing tool used and can be used with complex testing scripts. Conclusions There are no known frameworks or inbuilt support in existing load testing tools for guiding practitioners on applying Little's Law using output generated by tools. We address this significant gap by providing a framework which combines information from test scripts/reports and Little's Law to determine whether the results are valid. The provided implementation can be easily integrated with existing load testing tools.

Keyword:

Load testing

Performance testing

Little's law

Software quality

Little's law based validation framework for load testing

摘要

AI总结

期刊

机构

引用论文