Eventual Leader Election in Shared-Memory Dynamic Systems

In spite of the importance of dynamic distributed environments, e.g., SANs (storage area networks), multicore architectures, etc., we found very few proposals of models and protocols for implementing eventual leader election in these contexts. This abstraction is fundamental to ensure consistency and fault-tolerance. Most approaches for electing a leader consist of static systems, in which processes communicate by message passing, satisfying timing constraints. This paper presents an Omega (or Ω) class leader election service for a dynamic asynchronous system subject to crash failures, in which processes communicate through shared memory, following a memory access pattern, free of temporal requirements.