An agent based Distributed thermal balancing –Task migration

The system reliability, performance, cost, and leakage power in deep sub micrometer era have a significant impact by thermal hotspots and temperature gradients. As the system complexity increases, it is very difficult to perform thermal management in a centralized approach due to state explosion and the overhead of monitoring the entire chip. In this work, a framework for distributed thermal management in many-core systems where balanced thermal profile can be achieved by proactive task migration among neighbouring cores is proposed. The framework has a low cost agent residing in each core that observes the temperature and communicates with its nearest neighbour for task migration and exchange. Unnecessary migration are avoided by choosing only those migration requests that will not result in generating thermal emergency. Key Words— Distributed control, dynamic thermal balancing, multi-agent, task migration.