Research On Energy-efficient Scheduling Algorithms For Heterogeneous Computing Systems

The need of social life prompt the emergence and development of information science,while the development of information science on human society life,production and technological progress plays a huge role.A suite of distributed computing machines with varied computational capabilities which are interconnected by high speed links can be defined as a heterogeneous computing system.Heterogeneous computing systems have become an indispensable ingredient to human society life.However,the development of information and communication technologies brings people great convenience,while also many serious problems faced by people which urgently need to be addressed.Such as technical challenges caused by huge energy consumption,increasing electricity costs,tremendous additional costs including wider environmental impact on ecosystems and so on.So it becomes very necessary to study the prevalent heterogeneous systems,especially on the problem of energy consumption in systems.Reducing energy consumption can be achieved not only by hardware design but also software approaches.The Dynamic Voltage Scaling(DVS)and Dynamic Power Management(DPM)emerge as the times require,the DVS and DPM technologies have become the bridge of collaborating software and hardware to optimize energy consumption and have been used widely.As a core of a computing system,task scheduling is the key of managing the global resources and assigning resources.Binding a scheduling algorithm with the DVS or DPM has become an important method to reduce the energy consumption in systems.This paper proposes a DVS-based algorithm called Energy-Efficient Scheduling with Frequency Equalization(ESFE)for Heterogeneous computing systems.The ESFE approach aims to minimize the finish time as well as the overall energy consumption.First,the ESFE algorithm extracts the set of paths from an application.Then,the ESFE algorithm reconstructs the application based on the extracted set of paths in order to achieve a reasonable schedule.Finally,the ESFE algorithm adopts a progressive way to equalize the frequency of tasks in order to achieve the goals reducing the total energy consumption of systems.To validate the effectiveness of the ESFE algorithm,randomly generated applications and two real-world applications are examined in our experiments.The experimental results show that the ESFE algorithm outperforms two existing algorithms in terms of makespan and energy consumption.This paper presents a SLA-based Energy-Efficient Scheduling(SLAES)algorithm for parallel applications in the context of Service Level Agreement(SLA)on cloud computing environment.The SLAES approach aims to minimize the overall energy consumption while still meeting the certain performance goals based SLA.First,the SLAES algorithm comprehensively considers the high power processors to minimize the number of high power processors used.Then,the algorithm try to migrate some tasks from a high power processor to a low power processor for energy saving.Finally,the SLAES algorithm takes a more efficient way to assign the time slots among tasks based on the DVS technique.To demonstrate the effectiveness of the SLAES algorithm,randomly generated graphs and two real-world applications are tested in our experiments.The experiment results show that the SLAES algorithm can efficiently saves energy consumption of systems.