Group Communication: from Practice to Theory

Replication is a fundamental technique for achieving fault-tolerance in distributed systems. We start by presenting the two basic replication techniques, namely active replication and passive replication.

In a second step we introduce the group communication primitives that allow the implementation of active and passive replication: atomic broadcast and generic broadcast. We also clarify the advantage of group communication over quorum systems for implementing replication.

Finally we discuss issues related to the implementation of group communication: impossibility results, system models that allow the implementation of atomic broadcast/generic broadcast, and finally the most important algorithmic results.