面向能耗优化的云渲染系统任务调度策略

A Task Scheduling Strategy With Energy Optimization for Cloud Rendering Systems

针对云渲染系统中由于渲染节点与任务不匹配调度而带来的能耗浪费问题,提出一种通过任务调度方式来优化系统能耗的策略。为了形式化描述系统的整体能耗,综合考虑节点空闲能耗和任务运行能耗,建立渲染任务能耗模型;以降低系统总体能耗为优化目标,根据渲染任务之间无依赖性的特点,将任务调度序列拆分成子序列,利用模拟退火思想,通过优化子序列任务调度提高节点利用率、减少节点空闲能耗,以此降低系统全局任务的能耗;采用矩阵存储子序列任务的能耗,以空间换时间的方式降低策略的时间复杂度。实验结果表明:该策略在多渲染作业环境中能耗优化效果比先进先出算法提升了43.4%,比能耗感知的调度算法提升了6.7%,能够有效降低云渲染系统的总体能耗,同时具有良好的扩展性,使云渲染系统的能耗效率和整体性能得到提升。

A task scheduling strategy with optimized energy consumption for cloud rendering systems is proposed to solve the problem that the mismatching task scheduling on render nodes causes a great waste of energy consumption.A rendering task energy consumption model is presented to describe formally the overall energy consumption of a system and takes both the idle and the task running energy consumptions of each node into account.The optimization object is to reduce the overall energy consumption of the system,and the strategy divides the task scheduling sequence into subsequences based on the non-dependence characteristic among rendering tasks.The simulated annealing ideology is used to optimize the scheduling of the subsequence tasks,to improve the utilization ratio of the nodes and to reduce the idle energy consumption of nodes so that the energy consumption for the overall system is reduced.Moreover,the strategy adopts a way of space in time to reduce the time complexity by using a matrix to store the energy of subsequence tasks.Experimental results and comparisons with the FIFO algorithm and EMRSA（energy-aware MapReduce scheduling）algorithm in a multi jobs measurement show that the energy optimization performance of the proposed strategy has improved about 43.4%and 6.7%,respectively,that is,the proposed strategy effectively reducesthe overall energy consumption for cloud rendering systems.Moreover,the proposed strategy possesses better expansibility.It can be concluded that the proposed strategy can improve the energy efficiency and overall performance of cloud rendering systems.