Thermal-Aware Task Allocation and Scheduling for Heterogeneous Multi-core Cyber-Physical Systems

摘要

Over the past few years, the impact of heating on device failure rates has become increasingly important in cyber-physical systems. Such systems are often used in harsh environments and have weight, volume and power constraints which make it hard to dissipate heat economically and effectively. Thermal-aware task scheduling techniques are therefore called for to reduce heat stress on the computational platform. In this paper, we outline a thermal-aware task allocation and scheduling heuristic for use in computational platforms that feature heterogeneous computational cores. Such systems, consisting typically of some powerful, out-of-order, cores, together with simple, in-order, cores, allow the user a wider range of power-performance tradeoffs than traditional homogeneous multicore systems. We show how using fairly simple thermal-aware task allocation and scheduling principles results in a substantial enhancement in the expected lifetime of the system.