Task Allocation and Migration Algorithm for Temperature-Constrained Real-Time Multi-Core Systems

Temperature rise will affect the stability and performance of multi-core processors. A temperature-aware task scheduling algorithm for real-time multi-core systems, called LTEDF (Low Thermal Early Deadline First), is proposed in this paper. In LTEDF, a HCNF ( History Coolest Neighborhood First) task allocation algorithm is employed to balance the loads. When some cores are thermally saturated, task migration is performed to alleviate thermal saturation to get more uniform power density map. A thermal aware multi-core scheduling simulator(TAMSS) based on Hot Spot thermal model is implemented. TAMSS simulation results show that LTEDF algorithm can not only meet real-time guarantee, but also minimize the thermal penalty. Moreover, it can create a more uniform power density map than other thermal-aware algorithms, and significantly reduce thread migration frequency.