Building fault tolerant distributed systems using IP multicast

Our institute has been developing the only publicly available implementation of the Ada 95 Distributed Systems Annex for several years in strong collaboration with Ada Core Technologies. We are now extending this implementation far beyond the requirements of the Reference Manual by reducing the memory footprint and adding fault tolerance. This paper will present the progress made on the latter. So far, we have been using IP multicast and forward error correcting techniques successfully for data broadcasting as well as distributed semaphores for synchronization, to implement a one-writer, multiple-readers shared memory area. This allowed us to suppress the only single point of failure that existed in our implementation. We plan to use this model as a framework for construction of fault tolerant distributed systems.