Checkpointing with Multicast Communication

For long-running or large-scale distributed programs, the ability to provide software fault-tolerance via checkpointing is valuable. For scalable systems, multicast communication is becoming a predominant communication paradigm. While some aspects of consistency and channel state are the same for both unicast and multicast protocols, the implementation of checkpointing systems differ. This paper explores the problem of checkpointing in a multicast environment and introduces two checkpointing algorithms for such environments. The first algorithm is closely based on existing checkpointing algorithms. The second employs the multicast protocol to distribute checkpointing information efficiently.