Reconfigurable System-wide Monitoring Laying the Foundations for Autonomous Systems

Martin Schulz

Cornell University

Wednesday, March 17, 2004
10:00am - SENSQ 5317

Refreshments at 9:30am in SENSQ 5319

Abstract

As system complexity grows, the development of reliable and efficient systems requires that they be autonomous, self-aware, and self-adaptive.

The foundation for any adaptive system is its ability to monitor itself and to provide the obtained information to the respective system components as a basis for any adaptation. Current systems generally provide only limited monitoring capabilities, mostly in the form of few event counters. In this talk, I will introduce Owl, a monitoring framework designed to overcome these limitations by pervasively deploying programmable monitoring elements throughout the system. The design exploits FPGAs to realize hardware monitors located at event sources, such as memory buses. These monitors run and writeback results autonomously with respect to the CPU, thus mitigating the system impact of interrupt-driven monitoring or the need to communicate irrelevant events to higher levels of the system.

The framework is designed to be generally applicable to a broad range of monitoring goals. Owl's unique capabilities enable novel approaches to better system understanding: e.g., online trace compression; memory access histogram generation; and complex pattern recognition. Owl therefore provides a fundamental building block for any adaptive system requiring online information. This, in turn, paves the way for general autonomous computing systems.