Degree-Based Replica Placement Algorithms for P2P Data Grids

Replica placement, which decides when and where to create replicas, is an important technique for improving the efficiency of data grids. In this paper, degree is employed for replica replacement in P2P data grids. Two replica placement algorithms, Degree-based algorithm (DA) and Degree-frequency-based algorithm (DFA), are proposed. In DA and DFA we set a degree threshold p and a frequency threshold v to choose the candidate replica locations and a replica is placed in the candidate node which gives the minimum access cost. Simulation results showed that DA and DFA both can keep smaller Makespan and reduce the number of replicas compared with Simple, AlwaysReplica, and Ecomodel-Zipf. DFA proves to be superior to Frequency algorithms in reducing replica numbers, but DA can reduce Makespan more. In addition, we have studied the optimal parameters for the proposed algorithms.