Nonblocking Ordered Reliable Multicast In an Unreliable Distributed Environment

Processor coordination, in particular the consistent management of data replicated on different processors, is a major challenge for any distributed system. Coordination is particularly hard to achieve in the presence of failure events such as crashed of slow processors and communication links. In this work we explore a new solution to the coordination problem in an unreliable environment through the application of Nonblocking Ordered Reliable Multicast (NORM) protocols. Our work is motivated by the notion of asymetric safety which was first introduced in [11]. We show in this work that nonblocking ordered reliable multicast protocols with asymetric safety are feasible. We present efficient protocols that are fast, non-blocking and wait free. Furthermore, we illustrate the usefulness of such a protocol, by outlining a proctical system for providing high availability with load balancing by means of replicated load and availability data that is updated using our ordered reliable multicast