Thermal-Aware Global Real-Time Scheduling on Multicore Systems

Nathan Fisher, Jian-Jia Chen, Shengquan Wang, Lothar Thiele
2009 2009 15th IEEE Real-Time and Embedded Technology and Applications Symposium  
As the power density of modern electronic circuits increases dramatically, systems are prone to overheating. Thermal management has become a prominent issue in system design. This paper explores thermal-aware scheduling for sporadic real-time tasks to minimize the peak temperature in a homogeneous multicore system, in which heat might transfer among some cores. By deriving an ideally preferred speed for each core, we propose global scheduling algorithms which can exploit the flexibility of
more » ... core platforms at low temperature. Compared with load-balancing strategies, the proposed algorithms can significantly reduce the peak temperature by up to 30 • C to 70 • C for simulated platforms. aware scheduling algorithms focus on partitioned scheduling of periodic real-time tasks or a set of job instances without periodicity. Applying partitioned scheduling for real-time tasks in a multicore environment is often too conservative. The focus of this paper is obtaining results for thermal-aware scheduling under the global paradigm. This paper explores thermal-aware scheduling for sporadic real-time tasks to minimize the peak temperature in a homogeneous multicore system. As heat can transfer among cores and heat sinks, the cooling and heating phenomena is modeled by applying the Fourier's cooling model in the literature [10], [17], [23], [24] , in which the thermal parameters can be calculated by the RC thermal model. Although heat transfer is a dynamic process, it is not difficult to see that the temperature on a core is non-decreasing if the execution speed on a core is fixed. Moreover, it will end up with a steady state, in which the temperatures on all cores become steady. We show how to approximately minimize the peak temperature at the steady state. This paper proposes a two-stage approach. In the first stage, we derive the preferred speeds for execution to minimize the peak temperature under the necessary schedulability conditions of global scheduling. Then, in the second stage, we derive a proper speedup factor to satisfy the sufficient schedulability conditions of global scheduling. The proposed approach is quite general, and can be adopted for global scheduling algorithms that have both a necessary condition and a sufficient condition for the global schedulability of sporadic tasks, such as the global earliest-deadline-first (EDF) scheduling policy and the global deadline-monotonic (DM) scheduling policy. Furthermore, in our approach, we permit each core to have a potentially different speed than the other cores. To evaluate the effectiveness of the proposed algorithms, we use three multicore benchmarks with 4 × 1, 2 × 2, 3 × 2, and layouts for simulations. Compared with load-balancing strategies, the proposed algorithms can significantly reduce the peak temperature by up to 30 • C to 70 • C for simulated platforms. The rest of this paper is organized as follows: Section 2 shows the system model and problem definition. Section 3 presents how to derive the preferred speeds of cores for minimizing the peak temperature under the necessary schedulability conditions of global scheduling. Section 4 derives the feasible speed scheduling based on the preferred speeds. Section 5 presents performance evaluation over simulated multicore platforms.
doi:10.1109/rtas.2009.34 dblp:conf/rtas/FisherCWT09 fatcat:osek4tcxgbczzbxykjl5g4suyi