A synchronisation mechanism for replicated objects

Abstract

The HIDRA Concurrency Control (HCC) mechanism provides support for concurrency control in environments where the coordi-nator-cohort replication model is being used. This replication model allows the arrival of multiple invocations to different object replicas which serve locally those invocations and later make the appropriate checkpoints on the rest of replicas. The HCC uses a service serialiser object (SS) and a set of serialiser agents placed in each replica node. As a result, since the HCC components are replicated, this mechanism is also fault tolerant. Each invocation received by an object replica is processed by the SS which knows the invocations that are currently being processed. So, this agent is able to block or allow the execution of arriving invocations according to their conflicts with the currently active ones and the concurrency specification made when the object interface was declared.