A Copy Of The Data Grid Strategy Study

Data Grid has the abilities of distributed storage, which is able to contain data in various formats. Its appearance provides a good solution of magnanimous data processing.Replica is the one of the most important key technique in data grid. They are able to reduce the delay of data transfer and improve the efficiency of data access and processing. The existence of replica can also detract the access requirements of data on original node, balance the work load of nodes in grid and consequentially optimize the performance of the whole network. Besides, when some nodes are invalid, replica are also able to reduce the possibilities of data lost and enhance the stability.Current researches on replica in data grid are focus on the concentrative look up table model (e.g. Globus,Europe data grid project). In these models, the performances of key nodes limit the performance of whole system. In this thesis, we design a replica system without key nodes, which refers the current P2P technique. The main functions of this system include: replica creation and replacement, replica location and replica update. The slip windows technique is introduced in strategy of replica creation and replacement which is able to make immediate responses to the change of network; the module of replica location obtains the locations of replica according to data required. DHT technique can distract the information of replica into more than one node; the replica update module uses the loose asynchronous model to ensure the consistency between replicas. We simulate this system and compare it with some mature projects (e.g. Globus), the results show that this replica system has a better performance.