Implementation and Evaluation of a Fault-tolerant Mechanism on Network-wide Distributed Object-oriented Systems

摘要

This paper describes the implementation of a fault-tolerant mechanism in a CORBA- compliant object request broker (ORB). This mechanism manages the object replicas, thereby enabling non-stop facilities. It hides the internal structure of fault--tolerant objects and is implemented as an object adapter (OA) on the ORB, not as an ordinary object. Because it combines fault detection using the local timer of each node and maintenance for internal state consistency, the overhead is reduced. Evaluation of the mechanism's performance demonstrated that it is suitable for a distributed object-oriented processing environment be- cause it uses only low--overhead functions, such as message transfer, and not high-overhead functions, such as the dynamic invocation interface/dynamic skeleton interface (DII/DSI). This mechanism was developed for use in the distributed processing environment (DPE) of telecommunication networks in which continuous operation and connectivity to other systems are needed.