CAJ | 학술논문

在分布式计算系统下，网络中各个计算机终端节点的能耗管理是下一代网络分布式计算发展的瓶颈问题。传统的能耗管理调度方法采用单台计算机为模型进行分析，无法全面的权衡网络系统中各个节点的平均能耗，无法达到最优的能耗管理。提出一种基于多目标进化的分布式网络能耗管理办法，采用进化的思想，对系统对每个个体划分相应的等级，在此基础上，采用先种群等级分割，后个体等级排序的方法确定分割度，最后根据模型最优化的方向实现资源最优调度。通过2000个随机任务在10台分布式计算机上的运行情况进行实验，结果表明，采用本文的调度算法后，能耗大大降低，并且算法稳定，可靠，工程应用性很强。
재분포식계산계통하，망락중각개계산궤종단절점적능모관리시하일대망락분포식계산발전적병경문제。전통적능모관리조도방법채용단태계산궤위모형진행분석，무법전면적권형망락계통중각개절점적평균능모，무법체도최우적능모관리。제출일충기우다목표진화적분포식망락능모관리판법，채용진화적사상，대계통대매개개체화분상응적등급，재차기출상，채용선충군등급분할，후개체등급배서적방법학정분할도，최후근거모형최우화적방향실현자원최우조도。통과2000개수궤임무재10태분포식계산궤상적운행정황진행실험，결과표명，채용본문적조도산법후，능모대대강저，병차산법은정，가고，공정응용성흔강。
In distributed computing system, it was a serious problem of energy management for each node in the network sys-tem to develop. In traditional method, a single computer analysis model was used, but it was difficult to judge and weight en-ergy use of each node in the system and achieved the optimal management effect. A distributed energy optimal management module of network based on multi-object evolution was proposed, the developmental method was used to divide each node to appropriate grade, and then, the dividing degree was ensured with population firstly and individuality secondly, at last, according to the module, the optimal management was achieved. 2000 random tasks were taken in 10 computers to test the ability of energy management, and the result shows that with the proposed algorithm, the energy consumption is reduced greatly. The algorithm is stable and useful for engineering application.