Adaptive thermal-aware task scheduling for multi-core systems

• An online scheduling framework for multi-core real-time systems with thermal constraints and multiple quality of service levels is proposed.
• An integration of feedback controllers and thermal-aware multi-core scheduling algorithms is presented to manage on-line thermal dynamics.
• A service level controller is presented and used to cool the system by adjusting the quality of service of applications.
• Regarding uncertain thermal properties, the proposed framework exhibits stable results regarding overheating prevention and scheduling capability.

Thermal management is a challenging problem because of on-line thermal dynamics. An adaptive thermal-aware multi-core task scheduling framework based on run-time controllers is proposed in this paper to address the inter-core thermal effects and dynamic variations of task execution. In contrast to dynamic voltage scaling, the service rates of the tasks are adjusted in order to cool the system. Scheduling algorithms are used to prevent the system from overheating and maximize system utilization. This paper also evaluates the capability of the proposed framework through varying the workloads, ultimately demonstrating positive and stable performance.