The Research Of Replica Replacement Strategy In Data Grids

Now, the grid is the international frontier of research topics. With the development of Grid technology, the Grid begins to be used in various fields. The most important one is data grid; it is widely used in data-intensive industries. Data grid can access, storage, move or manage the data which is heterogeneous, distributed and mass. It is a very broad application prospects. The replica is an important part of the data grid, it can improve the performance of the system. However, with the increase of the gird node and running, the number of copies is much larger and the system seems bloated. At this time, too many copies bring the overhead which caused by the system that maintain consistency of replica. The overhead and the large storage space are more than the benefits of replica. Therefore, the research on the replica management technology is becoming more and more important.The replica management is made up of replica creation, replica selection, replica consistency and replica out. The solution of mass replica is removing. A good algorithm can improve the performance of network. Otherwise, the consumption of network bandwidth and the access latency of data file will increase. So, the deleting copy of the data grid can enhance the quality of the grid data service.The problem of the existing deletion algorithm is that the factors of affecting copy out are inadequate. They only consider the replica visited, do not concern the grid environment conditions, for example, network bandwidth and access delay, etc. The existing deletion algorithms appear to some extent inadequate in the grid environment that the gap of copy's size in it. In the paper, there is a comprehensive understanding of the background knowledge of the data grid. Base on the analysis of replica management correlation technique, there deeply study the deletion algorithm of replica.This article explains the factors affecting the deletion of replica. They are replicas'access time, replicas'access frequency and replicas'transmission cost. The article studies the balance problem of factors with each other. Then an improved replica's deletion algorithm is proposed. The algorithm not only considers the impact of elimination of the factors, but also introduces the dynamic adjustment factor. According to the size of replica, it can adjust dynamically the proportion of the various factors in the valuation model of replica. This algorithm can adapt to grid environment which the replica's size is different with eachother in it. It can reduce the error of copy-out and increase the effective utilization of network. We select the appropriate simulation software and verify the effectiveness of the improved algorithm. The grid environment adapted by the improved algorithm and the relation parameters are determined.