Sharing memory robustly in message-passing systems
Journal of the ACM (JACM)
RAMBO: A Reconfigurable Atomic Memory Service for Dynamic Networks
DISC '02 Proceedings of the 16th International Conference on Distributed Computing
How fast can a distributed atomic read be?
Proceedings of the twenty-third annual ACM symposium on Principles of distributed computing
Fault-tolerant semifast implementations of atomic read/write registers
Proceedings of the eighteenth annual ACM symposium on Parallelism in algorithms and architectures
Proceedings of the twenty-sixth annual ACM symposium on Principles of distributed computing
Hi-index | 0.00 |
Atomic (linearizable) read/write memory is a fundamental abstractions in distributed computing. Following a seminal implementation of atomic memory of Attiya et al. [6], a folklore belief developed that in messaging-passing atomic memory implementations "reads must write." However, work by Dutta et al. [4] established that if the number of readers R is constrained with respect to the number of replicas S and the maximum number of crash-failures t so that R V is constrained by V Quorum systems are well-known mathematical tools that provide means for achieving coordination between processors in distributed systems. Given that the approach of Attiya et al. [6] is readily generalized from majorities to quorums (e.g., [5, 2]), and that the algorithms in [4] and [3] rely on intersections in specific sets of responding servers, one may ask: Can we characterize the conditions enabling fast implementations in a general quorumbased framework? This is what we establish in this work.