基于资源调度的集群节能系统的设计与实现

集群计算机系统的运算性能跨入百万亿次、千万亿次时代,节能降耗已成为集群计算机系统必须面对的重要问题之一,本文从系统级节能的角度出发,结合神威高性能集群计算机系统的系统监测、作业管理、IPMI带外电源管理和TuxOnIce系统休眠技术,设计并实现了基于资源调度的集群节能系统,通过对空闲结点的关机或休眠,能够有效地降低集群系统空闲时的能耗,使神威高性能集群计算机系统成为真正的绿色计算机。

Design and Implementation Based on the Resource Scheduling for Power_Saveing Cluster System

When the cluster high performance computers (HPC) begin to develop 100Tflops or 1Pflops, How to save power is a very important question which we must affronted. This paper designs and implementations power_saving Cluster System based on the resource scheduling, this system attends the system class power_saveing, combines technology of the system monitor (SCSM) and job manager (SCJM) of Sunway cluster HPC, and IPMI's out-band power manager, and TuxOnIce's hibernation, which can reduce the power when the cluster HPC is idling and let Sunway cluster HPC to be really green HPC by shutdown or hibernating these idle computers.