CAJ | 학술논문

在分布式文件系统（ DFS）中，副本的数量通常预先配置，不能适应文件访问需求的动态变化。针对该问题，提出了基于热度分析的动态副本创建算法（ DRCA），优化副本的管理方式。动态副本创建算法通过分析文件在给定时间内的访问频率，推算文件的访问热度，并综合考虑统计周期、文件大小、工作环境等多种因素，按需动态地调整文件副本的数量，以减少文件访问的平均响应时间，提高数据服务性能。基于Hadoop的分布式文件系统（ HDFS）开发了DRCA模块，并进行了性能测试，结果表明DRCA提供数据服务的性能优于HDFS自带的副本创建算法。
재분포식문건계통（ DFS）중，부본적수량통상예선배치，불능괄응문건방문수구적동태변화。침대해문제，제출료기우열도분석적동태부본창건산법（ DRCA），우화부본적관리방식。동태부본창건산법통과분석문건재급정시간내적방문빈솔，추산문건적방문열도，병종합고필통계주기、문건대소、공작배경등다충인소，안수동태지조정문건부본적수량，이감소문건방문적평균향응시간，제고수거복무성능。기우Hadoop적분포식문건계통（ HDFS）개발료DRCA모괴，병진행료성능측시，결과표명DRCA제공수거복무적성능우우HDFS자대적부본창건산법。
In a distributed file system, the number of replicas is usually pre-configured which cannot adapt to the dynamic change of the file' s accessing need. For this problem, a Dynamic Replica Creation Algorithm ( DRCA) based on temperature' s analysis was proposed, which could optimize the management of replicas. Dynamic replica creation algorithm calculated file' s accessing temperature by analyzing file' s accessing frequency within the specified time, also considered multiple factors such as statistical period, file size and work environment, then adjusted dynamically the number of file' s replicas on-demand, thus reduced average response time of file' s accessing, and improved the performance of data service. The DRCA module was developed based on Hadoop Distributed File System ( HDFS) , and then its performance was tested. The results indicate that the DRCA ' s performance of providing data service is better than HDFS ' own replica creation algorithm.