An LIRS-Based Replica Replacement Strategy for Data-Intensive Applications

Abstract

Data Replication, creating identical copies of data on different sites geographically, is one of the effective optimization techniques for reducing data access costs and improving load balance among sites in data-intensive computing. However, due to the limitation of storage capacity of each site, an efficient replica replacement can make a significant difference in the performance and efficiency of replication technologies. In this paper, for the inability of LRU to cope with access patterns with weak locality, a novel replacement strategy called Low Inter-reference Recency Set (LIRS) was applied into replica replacement during data replication. And then, OptorSim was extended, and LIRS replica replacement algorithm was implemented. Finally, LIRS replacement strategy was comprehensively compared with other algorithms including LRU, MRU LFU and MFU by use of different scheduling algorithms and file access patterns, and the results show that LIRS replacement algorithm can improve the performance of the data access, especially under access pattern with weak locality.