Monitoring Membership Changes in a Fault-Tolerant Distributed System

The Totem protocol supports the maintenance of consistency of replicated information in fault-tolerant distributed systems by providing reliable totally ordered delivery of messages. The membership algorithm of Totem maintains a consistent view of processors in a local-area network and handles all aspects of reconnguration, including restarting of failed processors and remerging of partitioned networks. In this paper we describe a network monitor with graphical output that we have constructed for the Totem protocol development environment. The development environment executes the protocol object modules unmod-iied while simulating network communication, timing, and fault injection. The network monitor collects information from the processors within the development environment and displays global membership information and other data in an intuitive fashion.