Rearch Of Replica Management In Data Grid

Grid is an emerging technology, it connects the numerous information and resourceisland, So that the people’s work and life are more convenient. Computing grid is aplatform of compute intensive application, and Data grid derives from computing grid, itis a distributed platform of data intensive application and computation intensiveapplication, but not have strict boundaries with computing grid. Data grid is a gridplatform which takes data as major resources. The core technologies of data grid is replicamanagement strategy, the effectively strategy of replica management can reduce thetransmission delay and network bandwidth consumption etc.Firstly, the current status of the replica management strategies are analyzed, for thecurrent efficiency of data management strategy is not high, poor performance and othercharacteristics, the three aspects of replica selection, replica replacement and replicacreation are analyzed and studied. Reasonable use of the grid environment and makeeffective replica creation can improve the performance of data grid system, the replicalocation and selection can find the best copy, replica replacement is the core technology ofreplica management. Because of the space of storage elements in each node is limited, so agood replica replacement strategy can improve the job execution speed, the utilization rateof storage element and reduce the consumption of bandwidth. Replica consistencytechnique effective solved the consistency problem of two or more replicas.Secondly, the replica creation, replica selection of DHRA replica managementstrategy are reached, their shortcomings are analysis. Considering the network bandwidth,file size and the number of copies of the request, improved replica selection algorithm andimproved replica creation algorithm are proposed.Thirdly, the replica replacement of DHRA replica management strategy is reached,and the shortcoming of only consider the historical access time in replica replacement isanalysed.Then the replica creation, replica selection and replica replacement of DHRA replicamanagement strategy are reached, and proposed three new algorithms for these threeaspects. Improved replica replacement algorithm based on replica weight and replica cost is proposed. A new replica management strategy is proposed according to thesealgorithms.Finally, using the data grid simulation tool Optorsim, doing a simulation experimentsof LWLC replica management strategy. And the simulation experiments of DHRA replicastrategy, LRU replica management strategy and LFU replica management strategy aremade. The experimental results were compared and analysed, the results demonstrate theeffective of LWLC replica management strategy.