A Sensor-Based Approach to Symptom Recognition for Autonomic Systems

The increased complexity of today´s distributed, composite, Web-based systems presents difficult and unique systems management problems. The way these systems interact, and the fact that they often span organizational boundaries, render them difficult to monitor and manage. Autonomic computing has emerged as a promising approach to the management of complex systems. A key to realizing fully autonomic systems is the development of monitoring tools that provide the controllers with adequate and meaningful performance information, especially the identification of symptoms that indicate potential underlying problems. We present an event-driven sensor approach to a monitoring system whereby a hierarchy of dedicated, simple sensors monitors and correlates low level events into a meaningful representation of the system performance that can be used for problem determination. Our approach utilizes the OASIS Web services distributed management (WSDM) standards.