Dakai Zhu

Power management has become a very important research area and various approaches have been proposed. As an energy-efficient architecture, chip multiprocessor (CMP) has been widely adopted by chip manufacturers. In this paper, we study power management schemes for real-time systems on block-partitioned multicore platforms, where the processing cores are grouped into different blocks and cores on one block share the same power supply voltage (thus have the same frequency). We evaluate the energy efficiency of different block configurations. Simulation results show that block-partitioned CMP has its inherent advantages to fulfill the goal of power efficiency and low design complexity for future CMPs.

Dynamic voltage and frequency scaling (DVFS) has been widely used to manage energy in real-time embedded systems. However, it was recently shown that DVFS has direct and adverse effects on system reliability. In this work, we investigate static and dynamic reliability-aware energy management schemes to minimize energy consumption for periodic real-time systems while preserving system reliability. Focusing on earliest deadline first (EDF) scheduling, we first show that the static version of the problem is NP-hard and propose two task-level utilization-based heuristics. Then, we develop a job-level online scheme by building on the idea of wrapper-tasks, to monitor and manage dynamic slack efficiently in reliability-aware settings. The feasibility of the dynamic scheme is formally proved. Finally, we present two integrated approaches to reclaim both static and dynamic slack at runtime. To preserve system reliability, the proposed schemes incorporate recovery tasks/jobs into the schedule as needed, while still using the remaining slack for energy savings. The proposed schemes are evaluated through extensive simulations. The results confirm that all the proposed schemes can preserve the system reliability, while the ordinary (but reliability-ignorant) energy management schemes result in drastically decreased system reliability. For the static heuristics, the energy savings are close to what can be achieved by an optimal solution by a margin of 5 percent. By effectively exploiting the runtime slack, the dynamic schemes can achieve additional energy savings while preserving system reliability.

Considering the impact of the popular energy management technique Dynamic Voltage and Frequency Scaling (DVFS) on system reliability, the Reliability-Aware Power Management (RA-PM) problem has been recently explored to save energy while maintaining system reliability. In this work, focusing on Rate Monotonic Scheduling (RMS) policy, we study static RA-PM schemes for periodic realtime tasks. After showing the intractability of the problem, we focus on two widely-known feasibility tests for RMS (namely, the Liu-Layland bound and Time Demand Analysis) and propose a number of heuristics based on the priority-monotonic speed assignment. The heuristics are evaluated through extensive simulations.

Cluster scheduling, where processors are grouped into clusters and the tasks that are allocated to one cluster are scheduled by a global scheduler, has attracted attention in multiprocessor real-time systems research recently. In this paper, assuming that an optimal global scheduler is adopted within each cluster, we investigate the worst-case utilization bounds for cluster scheduling with different task allocation/partitioning heuristics. First, we develop a lower limit on the utilization bounds for cluster scheduling with any reasonable task allocation scheme. Then, the lower limit is shown to be the exact utilization bound for cluster scheduling with the worst-fit task allocation scheme. For other task allocation heuristics (such as first-fit, best-fit, first-fit decreasing, best-fit decreasing and worst-fit decreasing), higher utilization bounds are derived for systems with both homogeneous clusters (where each cluster has the same number of processors) and heterogeneous clusters (where clusters have different number of processors). In addition, focusing on an efficient optimal global scheduler, namely the boundary-fair (Bfair) algorithm, we propose a period-aware task allocation heuristic with the goal of reducing the scheduling overhead (e.g., the number of scheduling points, context switches and task migrations). Simulation results indicate that the percentage of task sets that can be scheduled is significantly improved under cluster scheduling even for small-size clusters, compared to that of the partitioned scheduling. Moreover, when comparing to the simple generic task allocation scheme (e.g., first-fit), the proposed period-aware task allocation heuristic markedly reduces the scheduling overhead of cluster scheduling with the Bfair scheduler.

While Dynamic Voltage Scaling (DVS) remains as a popular energy management technique for real-time embedded applications, recent research has identified significant and negative impact of voltage scaling on system reliability. For this reason, a number of reliability-aware power management (RA-PM) schemes were recently proposed to preserve the system reliability when DVS is used. In this paper, we propose a new approach, called the shared recovery (SHR) technique, to minimize the system-level energy consumption while still preserving the system's original reliability. The main idea of the SHR technique is to avoid the offline allocation of separate recovery tasks to the scaled tasks by assigning a global/shared recovery block that can be used by any task at run-time. Our simulation results show that, compared to the existing RA-PM schemes, our scheme can achieve up to 35% energy savings. Further, this performance is shown to be comparable to the maximum energy savings that can be achieved by any algorithm. Interestingly, our extensive evaluation indicates that SHR offers also non-trivial gains over the previous algorithms on the reliability side. Further, a dynamic extension is proposed to improve energy and reliability management at run-time by reducing the size of the recovery block and re-using the slack that arises from early completions.

Recently, the negative effect of the popular power management technique Dynamic Voltage and Frequency Scaling (DVFS) on the system reliability has been identified. As a result, various reliability-aware power management (RAPM) schemes have been studied for uniprocessor real-time systems. In this paper, we investigate global scheduling-based RAPM (G-RAPM) schemes for a set of frame-based real-time tasks running on a homogeneous multiprocessor system. An important dimension of the problem is how to select the appropriate subset of tasks for energy and reliability management (i.e., schedule a recovery for each selected task and scale down their executions). We show that making this decision optimally (i.e., the static G-RAPM problem) is NP-hard. Then we propose two efficient G-RAPM heuristics, which rely on local and global task selections, respectively. Moreover, to reclaim dynamic slack generated at runtime, we extend the slack-sharing based global dynamic power management scheme to the reliability-aware settings. The proposed schemes are evaluated through extensive simulations. The results show that our static G-RAPM heuristics can preserve system reliability while achieving significant energy savings (within 3% of an upper bound for most cases). Moreover, G-RAPM with global task selection provides better opportunities for dynamic slack reclamation and up to 15% more energy savings can be obtained at runtime compared to that of local task selection.

Abstract— The Dynamic Voltage Scaling (DVS) technique is the basis of numerous,state-of-the-art energy management schemes proposed for real-time embedded systems. However, recent research has illustrated the alarmingly negative impact of DVS on task and system reliability. In this paper, we consider the problem of processing frequency assignment to a set of real-time tasks in order to maximize the overall reliability,