ISIS is a system developed out of a study of fault tolerance in distributed systems. The system implements a collection of techniques for building software for distributed systems that performs well, is robust despite both hardware and software crashes, and exploits parallelism. The basic approach is to provide a toolkit mechanism for distributed programming, whereby a distributed system is built by interconnecting fairly conventional nondistributed programs, using tools drawn from the kit. Tools are included for managing replicated data, synchronizing distributed computations, automating recovery, and dynamically reconfiguring a system to accommodate changing workloads. ISIS has become very successful: hundreds of companies and Universities currently employ the toolkit in settings ranging from financial trading floors to telecommunications switching systems.
The ISIS project has moved from Cornell University to Isis Distributed Systems a subsidiary of Stratus Computer, Inc.
The Horus project was originally perceived as a redesign of the Isis group communication system, but has evolved into a general purpose communication architecture with advanced support for the development of robust distributed systems in wide variety of settings. A first version of the Horus software is available to research institutions at no cost.
Some links that may be of interest: