Evaluating and Comparing the Impact of Software Faults on Web Servers

The impact of software faults present in components to the larger system is currently a relevant and still open research topic. Web-based applications are simultaneously a relevant type of system for our society and are typically exposed to many software components in the server side. The impact of faults in these components to the web servers is an important aspect when evaluating the dependability properties of the entire web-serving system. This paper proposes an experimental approach to evaluate and compare the impact of software faults present in web applications on typical web servers. This approach consists in emulating realistic software faults in server-side web applications and monitoring the behavior of web server from both the server side (e.g., resource consumption) and the client side (e.g., response time, response correctness) perspective. We exemplify our methodology in case studies using three different servers and a realistic e-commerce web application and show that software faults existing in server side components can indeed affect the web server in a quantifiable manner which allow us to use our methodology for comparative purposes towards benchmarking and selecting the most robust web server.