Thermal-aware power-efficient deadline based task allocation in multi-core processor

Abstract

This paper address the power-efficient task allocation problem in a heterogeneous multi-core processor for thermally challenged tasks with real-time constraints. A thermally challenged task is one which, if executed on a core at the maximum possible speed, resulting the core temperature exceeds a given safe temperature limit. Our problem formulation is based on the well-praised thermal models HotSpot-4, leakage power and delay models with DVFS (dynamic voltage and frequency scaling). The solution to this problem is not so straight forward due to the fact that there is a nonlinear-circular dependency between the overall power dissipation and the temperature. It becomes further complicated with the consideration of a crucial parameter, task's deadline. We have solved this problem with a novel idea of deadline based task allocation scheme (DBTAS) that takes care of the permissible thermal dissipation, power consumption and speed of the processor. We have included suitable numerical test cases to demonstrate the efficacy of the proposed scheme for task allocation.