Skip to Main Content
Checkpoint and Message logging as an effective and transparent recovery method usually are implemented in LAN. With the development of scale in distributed system, computing nodes are located differently in geography. Fault tolerance and scalability problems are a real challenging work in wide area environment, because the length of wrapped message is increasing based on the node numbers. Some improved methods have brought false dependable problem and caused the system to roll back additionally. We proposed a scheme in which besides exchanging inter-domain messages by gateway two-dimensional state vector is used to divide and track the dependency. By the means the false dependency avoided. The proposed scheme can not only employ hierarchy gateways but also composite recovery protocols, so it can adapt to complexity situation of distributed system in wide area network.