Abstract: The traditional Master-slave model of volunteer computing is limitation due to the compositional nature of volunteer computing. Considering the volunteer nodes are usually from different locations, a Network-aware based model for fault tolerant volunteer computing was proposed. This model considers the network of the volunteer nodes and partilions the nodes into different subsets, based on this model volunteer computing system can tolerant some fault caused by network delay or timeout. In addition, the overlapping strategy was improved and it was applied to the fault tolerant volunteer computing. I}he results of experiments show that the network-aware model and improved overlapping strategy can improve the reliability and performance of the volunteer computing.

Key words: Volunteer computing, Network-aware, Fault tolerance, Overlapping strategy