Storm平台非关键路径电压调控节能策略

Energy-efficient strategy for non-critical path by voltage regulation in Storm

针对流式大数据计算平台的高能耗问题,提出一种基于Storm平台的非关键路径电压调控节能策略.首先根据数据流处理特点建立有向无环图、关键路径及电压调节模型,确定位于拓扑执行非关键路径上的工作节点CPU使用率及数据传输量阈值;其次根据拓扑执行非关键路径上工作节点的不同情况,对工作节点内存电压进行调节,当工作节点的CPU使用率及数据传输量低于阈值时降低其内存电压,当工作节点CPU使用率及数据传输量高于阈值时则升高其内存电压,由工作节点内存电压确定系统功率;最后通过评估性能与能耗的关系验证算法的可行性.实验结果表明,系统实施节能策略后比原系统降低了约34.5%的能耗,且位于拓扑执行非关键路径上的节点越多节能效果越好.

In order to solve the problem of high energy consumption of big data stream computing platform,an energy-efficient strategy of non-critical path voltage regulation based on storm platform is proposed.Firstly,models of directed acyclic graph,critical path and voltage regulation model are established according to the features of data stream processing,the appropriate threshold values fit for the CPU utilization and the volume of transmitted data of work node executing on the non-critical path of a topology are determined.Secondly,the DRAM voltage of the work node is adjusted according to the different conditions of the work node in the non-critical path.The DRAM voltage of the work node is reduced when the CPU utilization and the volume of transmitted data are lower than the threshold value,and the DRAM voltage is increased when the CPU utilization and the volume of transmitted data of the work node are higher than the threshold value,and the system power is determined by the DRAM voltage of work node.Finally,the feasibility of the algorithm is verified by evaluating the relationship between ratio of performance and energy consumption.The experimental results show that the system which using energy-efficient strategy is more effective than the original system by 34.5%,and the more nodes located on the non-critical path of a topology,the better the energy efficiency.