Improved Strategy For The Energy-efficient Scheduling Algorithm Based On Heterogeneous Computing Systems

Recently,with the computing power of heterogeneous computing systems,energy consumption has a direct impacted on the development and application of heterogeneous computing systems,energy savings has become an important factor to be considered in heterogeneous computing systems.Therefore,for heterogeneous computing systems,increasing attention has been directly towards the energy efficient scheduling algorithm.Energy efficient task scheduling based on heterogeneous computing systems is proven to be a typical NP-complete problem.And after considering energy consumption,the problem becomes more complex;therefore,the energy-efficient scheduling algorithm based on the list algorithm is adopted frequently.However the task priority is calculated as the b-level value in the task priority calculation stage of the present algorithm,without considering other priority calculation method;in the task allocation phase the existing algorithms only balance schedule lengths and energy savings from the perspective of a single task scheduling,there is a strong locality;and cannot be automatically adjusted according to the application requirements.In response to these problems,improved strategies are proposed based on previous works in this paper.Details are as follows:To settle the insufficient of the research in the task priority calculation and task allocation of the existing energy-saving scheduling based on the list algorithm,firstly,this paper analyzes and compares the three task priority calculation method used in heterogeneous computing systems,from which to explore the optimal method;in task allocation phase,through the analysis of the impact of the implementation of the single task on the entire system,this paper designs a new objective function.On this basis,we propose an improved list-based energy-efficient scheduling algorithm.The improved algorithm can balance schedule lengths and energy consumption more efficiently.On account of the different schedule lengths requirements of different application,this paper analyzes the definition and principles of adaptive control and its application in energy-saving scheduling,combined with scheduling result mainly from the objective function in the list algorithm,study the relationship between the proportions of task completion time and scheduling results,we designed an adaptive energy-efficient scheduling control algorithm based on variable weights.Weight-based energy-efficient scheduling adaptive control algorithm can reduce the system power consumption as more as possible on the premise that the application requirements can be meet.To validate the algorithm's effectiveness,a simulation system is designed to the current energy-efficient task scheduling models based on the heterogeneous computing systems.The strategies proposed in this paper and exiting algorithms are simulated under the same tasks,power and processing units' parameters,and evaluate the strategies from the energy consumption and the schedule lengths.Theoretical analysis and experimental results show that energy-efficient task scheduling proposed in this paper can balance schedule lengths and energy savings more effectively;with three different methods of computing task priority,the algorithm's performances are not very different;energy-efficient scheduling adaptive control strategy proposed in this paper can significantly reduce energy consumption on the premise that the application requirements is meet.