Research Of Replica Model And Data Moving Strategy In Data Grid

DataGrid aims to integrate huge amount of heterogeneous and self-autonomous resources which are dispersed on the network and supply computing service, storage service and information service to the end users by the Grid Technology. A suitable replica policy is very important to the efficiency and reliability of the DataGrid, meanwhile, the ability to move the data with the distribution of the computing resource to meet the data needs of the jobs running on the computing resources and the needs of users in the other domain is also very important.Data Service is one of the kernel services in the ChinaGrid Support Platform, which is responsible for the management of underlying storage resources and user's virtual space in the ChinaGrid, providing high-ability data access services for grid users and the job execution engines such as General Running Service, Grid Parallel Programming, etc.The key standards to evaluate the data management service of the Grid Platform are the reliability, availability of the Grid System and the job's latency of the job's execution engines. By constructing a suitable replication model, we can still access the data when some data resources had corrupted, and the system can regenerate the data from the remaining replication. The reliability and availability can be maintained in this way. The job that the user submitted to the grid uses both the computing resources and storage resources in the grid environment. The data service of the grid should track the computing resources to minimize the transfer latency of the files that the job needed. By studying the dynamic data moving strategy, the data in the grid can flow to the hot spot of the computing resources. The jobs that are submitted to the grid can have a shorter latency. These two aspects complement with each other and ensure the reliability, availability and performance of DataGrid together.The real system tests shows that the replica model has a complete function implementation in the CGSP Data Service and can improve the efficiency of it. Meanwhile, the simulation tests shows that the introduction of the data moving strategy based on the replica model can improve the running efficiency of the jobs in the Grid.