Research On Dynamic Energy-efficient Scheduling For Cluster System

Nowadays, the distributed computer clusters becoming more and more widely used,the energy consumption of the computer cluster has become a hotspot of academicresearch in recent years. Comparing with the energy-saving technologies of large-scalecomputer clusters, the energy-saving strategy of other scale cluster is weak. The researchon energy-saving method of these clusters is of great significance.By building a typical distributed computing environment and analyzing the energyconsumption of computer components and inter-nodes under different workloads,designed a two-level dynamic energy-saving scheduling algorithm. At the CPU level, thealgorithm analyses the utilization of CPU in clusters, set the schedule strategy and theCPU frequency adjust method, use CPUfreq module to achieve the adjustment of the CPUfrequency. At system level, the algorithm analyses the cluster system history log, design asleep/wake strategy with self-learning ability, through reasonable sleep/wake, and achievea dynamic balance between workload performance and energy consumption.The experimental results show that the CPU frequency adjust strategy can reduceabout10%energy consumption under best condition, and the sleep/wake strategy canreduce about18%energy consumption under distributed environment. At the last of thetest, we measure the wake time and wake energy consumption of different kinds ofcomputers, putting out advices of CPU frequency adjustment strategy and the goodnessand weakness of the sleep/wake strategy, giving out an energy-saving principle.