Temperature-aware Real-Time Task Scheduling For Heterogeneous Multiprocessor Systems

With the great advance in information technology, the computer with high data processing ability is desirable. Multiprocessor technology is one common solution to handle this challenge, and its key part is task scheduling technique. As the increasing development of multiprocessor technology, the microprocessors in the system become diversity and heterogeneous. As a result, to deal with the heterogeneity of microprocessors, task scheduling problems are hence more complicated.This paper proposes a temperature-aware task scheduling algorithm to optimize energy consumption and peak temperature of a heterogeneous multiprocessor real-time system. At first, for the heterogeneous multiprocessor systems where the frequency of each processor is fixed, this paper introduces a concept of energy density and builds a minimal energy density cost model. Based on the model, this paper proposes a real-time task allocation scheme to reduce system energy consumption by greedily binding tasks and processors. Furthermore, for the heterogeneous multiprocessor systems where the frequency of each processor is variable, this paper proposes a time point based heuristic searching algorithm to reduce the system energy consumption under the constraint of peak temperature. The searching method utilizes a well-designed estimation function that can guarantee an energy efficient solution among the task allocation solution space.The simulation results show that the proposed temperature-aware energy efficient task scheduling algorithms are effective in reducing energy consumption and improve quality of service. Compared to the benchmarking method RMFF[1], the proposed scheme can reduce the energy consumption in the range of 21-28%, and lower the peak temperature in the range of 7-10℃. In particular, when the task deadline miss ratio is low, the system can achieve a high quality of service.