Impossibility of distributed consensus with one faulty process
Journal of the ACM (JACM)
Unreliable failure detectors for reliable distributed systems
Journal of the ACM (JACM)
DSN '02 Proceedings of the 2002 International Conference on Dependable Systems and Networks
Eva: An Event-Based Framework for Developing Specialized Communication Protocols
NCA '01 Proceedings of the IEEE International Symposium on Network Computing and Applications (NCA'01)
Impact of a Failure Detection Mechanism on the Performance of Consensus
PRDC '01 Proceedings of the 2001 Pacific Rim International Symposium on Dependable Computing
Hi-index | 0.00 |
The design of most distributed consensus protocols doesnot take into account the fact that, over time, the differentprocesses running the protocol are subject to a varyingavailability of processing and communication resources.Because of that, performance bottlenecks are possible toarise in those executions on which a process that playsa key role in the protocol is executing on a slower processor,or experiences long delays in the communicationwith other processes. In a previous work we have proposedthe abstraction of a slowness oracle that allows a classof consensus protocols, designed for the asynchronous distributedsystem augmentedwith unreliable failure detectors,to adapt themselves to this varying availability of resources,minimizing performance bottlenecks. In this paper we analysethe performance of several implementations of suchadaptive consensus protocols. Their performance is comparedagainst that of a non-adaptive protocol via both simulatedscenarios as well as experiments conducted at ourlaboratory. From the results attained, we conclude that themore accurate is the slowness oracle the better is the performanceof the adaptive protocols that uses it. Nevertheless,even very naive implementations of a slowness oracleare suf.cient to yield better performance when system loadis not homogeneous.