Energy efficient scheduling for real-time embedded systems with QoS guarantee

Abstract

While the dynamic voltage scaling (DVS) techniques are efficient in reducing the dynamic energy consumption for the processor, varying voltage alone becomes less effective for the overall energy reduction as the static power is growing rapidly. On the other hand, Quality of Service (QoS) is also a primary concern in the development of today's pervasive computing systems. In this paper, we propose a dynamic approach to minimize the overall energy consumption for soft real-time systems while ensuring the QoS-guarantee. The QoS requirements are deterministically quantified with the window-constraints, which require that at least m out of each non-overlapped window of k consecutive jobs of a task meet their deadlines. Necessary and sufficient conditions for checking the feasibility of task sets with arbitrary service times and periods are developed to ensure that the window-constraints can be guaranteed in the worst case. And efficient scheduling techniques based on pattern variation and dynamic slack reclaiming extensions are proposed to combine the task procrastination and dynamic slowdown to minimize the energy consumption. In contrast to the previous leakage-aware dynamic reclaiming work which never scales the job speed below the critical speed, we will show that it can be more energy efficient to reclaim the slack with speed lower than the critical speed when necessary. Through extensive simulations, our experiment results demonstrate that the proposed techniques significantly outperformed the previous research in both overall and idle energy reduction.

On the Performance of Greedy Algorithms for Power Consumption Minimization

“…On the practical side, further work could be devoted to conducting experiments with a more complicated power model, that would include static power in addition to dynamic power (see for example the model for the Intel Xscale [15], detailed in [16], [17], [18]). With such a model, the "natural" greedy algorithm would assign the next job to the processor that minimizes the increment in total power.…”

Improving schedulability and energy efficiency for window-constrained real-time systems with reliability requirement

“…The mandatory jobs are the jobs that must meet their deadlines in order to satisfy the ðm; kÞ-constraints, while the optional jobs can be executed to further improve the quality of the service or simply be dropped to save computing resources. In [28], the same strategy was adopted to satisfy the window-constraints for real-time systems.…”

“…Some other studies [29,2,28,11] have also been conducted related to reducing the energy consumption for real-time systems with QoS constraints based on mandatory/optional job partitioning strategies. In [29,28], approaches adopting dynamic mandatory/optional job patterns were introduced to reduce the energy consumption for real-time systems with QoS-guarantee.…”

“…The mandatory jobs are jobs that must meet their deadlines in order to satisfy the ðm; kÞ-constraints, while the optional jobs can be executed to further improve the QoS or simply be dropped. In [28] a similar partitioning strategy was adopted to deal with the problem of scheduling with window-constraints guarantee for task sets with arbitrary execution times and periods.…”

Energy-Aware Mappings of Series-Parallel Workflows onto Chip Multiprocessors

“…CMP configuration. For processor speeds and power consumption, we use the model of the Intel Xscale [23], following [13,11,36]. There are five speeds for each core: s u,v = (0.15, 0.4, 0.6, 0.8, 1) GHz, with power consumption P (comp) su,v = (80, 170, 400, 900, 1600) mW .…”