S.Ghemawat et al [7] discussed about chunk size of a file that store in storage devices. It suggest 64 mega bytes size chunk. That decrease number of meta data that store in a master server. That increase master server efficiency. Data flow is another important point of the server. Network bottleneck another issue that face any network. Specially in data mining data flow is continuously happen. Parse data to master server in closest path is a solution for that. Distance from peer to peer calculate by ip address (internet protocol address). Data recovery and replication approaches are based on two main categories. Chunk replication and Master replication make system reliable. Chunk replication done by priority of data set. High priority data save in many peer machines. Low priority data save in less number of peer machines. Master recovery done by two shadow servers.