Agent Fault Tolerance Using Group Communication

摘要

This paper describes an agent fault-tolerance protocol called SG-ARP that allows mobile agents to execute correctly despite node or communication failures. SG-ARP uses three fault-tolerance techniques: group communication, process checkpointing and recovery, and TCP. It does not require any data transfer or network connections over the Internet, is COTS-based, has low performance overhead, does not require any user involvement, and preserves the output commit property.