This research attempts to provide fault tolerance for distributed systems using the concept of resilient forward recovery. The concept uses replication to achieve fault tolerance, and goes further to dynamically regenerate replication in response to failure. The concepts are incorporated through a library technology that is application independent. Replicated processes are created transparently to the application layer and are formed into a process group. Each group has a group manager whose job is to coordinate liveness checking and conduct system recovery. Initially, the group manager will be created. Subsequently, the manager will create and manage other processes in the group. Our library model hides the details of communication protocols required to achieve dynamic replication and reconfiguration in distributed applications. A reliability assessment for the library prototype is also presented. This paper provides a status report on our process in developing the concepts.
展开▼
