Graceful degradation in server overload using adaptive content
Henri Naccache · 2008
Current web application-level Quality of Service (QoS) frameworks resort to refusing requests when met with overload. The proposed self-healing adaptive content framework (SHAC) will allow dynamic web-based applications to adapt their output in order to minimize the effects of flash crowds and server overload without refusing requests. A methodology to create a SHAC system for existing web applications is presented. An analytical Queuing Network model is created to predict the performance impact of the framework on existing systems. A PID controller-based autonomic manager is implemented to maintain response times below that of a preset service level agreement. Evaluation of the framework is shown through implementations in web services, a portal, an AJAX-driven systems and a blog server along with significant performance testing and QoS analysis.