Using Proportional-Integral-Derivative Control in Self-Healing Adaptive Content Systems

Database-driven web applications have become the norm, especially in-service-oriented applications. As network bandwidth has increased the bottleneck in web application survivability has moved to the server and application level. With social bookmarking sites increasing in popularity, flash crowds are becoming more prevalent leading to higher than expected request loads. In this paper we describe the self-healing adaptive content (SHAC) system used to prevent server overload at the application level. We respond to requests with multiple resolutions of the textual content, using lower resolution requests to reduce the service demand on server hardware. We present a novel proportional-integral-derivative (PID) controller-based autonomic manager that is seeded from a queueing network (QN) model in order to best manage response times. In this paper we describe the methodology used to implement the SHAC system and evaluate the impact of the system on a popular open source blog server.