Fault-tolerant remote procedure calls

Distributed systems are potentially more reliable and resilient than centralized ones. They provide resource sharing among the network. Remote procedure call (RPC) appears to be an useful paradigm for providing communication across a network between programs written in a high-level language. It is widely used in various distributed systems. However, remote transactions may fail if the node on which a remote procedure is executing fails. The authors present a scheme to make RPC fault tolerant to hardware failures. Calling information is logged onto a stable storage for recovery usage. RPC services are registered at decentralized name servers. A commit protocol and a recovery scheme are developed for keeping the system in a consistent state. The authors also provide a new stub generator that makes RPC fault tolerant without changing the source codes.