Usage-based statistical testing of web applications

  • Authors:
  • Jianhua Hao;Emilia Mendes

  • Affiliations:
  • Provenco Retail Automation, Auckland, New Zealand;The University of Auckland, New Zealand

  • Venue:
  • ICWE '06 Proceedings of the 6th international conference on Web engineering
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

The large growth in the number of large Web applications currently developed brings concerns related to their quality, and more specifically their testing and reliability. Web application testing is still in its infancy and relies mostly upon traditional software coverage testing processes, which are highly impractical. Recent studies have looked at usage-based statistical models for testing Web applications. One of these studies, conducted by Kallepalli and Tian [8], used Unified Markov Models (UMMs), built from Web server access logs, as basis for test case selection. In addition, server error logs were also used to measure a Web application's reliability, and consequently to investigate the effectiveness of UMMs as a suitable testing mechanism. This paper describes two experiments that replicated Kallepalli and Tian's work. Our results showed that, in contrast to [8], multiple set UMMs were needed for trustworthy test case generation. In addition, our reliability assessment corroborated results from [8], confirming that UMMs seem to be a suitable testing mechanism to use to test Web applications.