Strong and weak virtual synchrony in Horus
SRDS '96 Proceedings of the 15th Symposium on Reliable Distributed Systems
Hi-index | 0.00 |
Large-scale systems spanning geographically distant sites are potentially appropriate environments for distributed applications supporting collaboration. In this paper, we examine the possibility of using such systems as repositories for replicated files to facilitate low-latency data sharing. Asynchrony in communication and computation, complex combinations of site and communication failures, and in particular, network partitions that characterize these systems make the design of algorithms to operate on them a difficult task. We show that view-synchronous communication is not only an appropriate conceptual model for reasoning about large-scale distributed systems, it is also an effective programming model. We support these claims by developing algorithms for managing replicated files with one-copy serializability as the correctness criteria.