\the Totem Single-ring Ordering and Membership Protocol," Acm Transactions on 6 Conclusion


membership changes in a fault-tolerant distributed system," In Proceedings of the Communication protocols that provide reliable ordered message delivery are important for distributed systems. The lack of a controlled environment impedes the testing, de-bugging and validation of such protocols. This motivation led to our creation of the Totem protocol development environment. The discrete-event simulator in this development environment allows a developer to inject faults in a simulated environment and to obtain an accurate representation of how the protocol handles these faults. As our target protocols matured with new capabilities that provide reliable ordered delivery to more than one interconnected LAN, the simulator was unable to continue providing the service for which it was created. The simulator was thus enhanced with the virtual gateway mechanism. It allows us to link concurrent simulations of single LANs on individual processors into one large distributed simulation of interconnected LANs on multiple hosts. To eliminate the possibility of deadlock, we devised a congestion control scheme that limits the number of messages in the system overall. The reened simulator and its potential to be adapted to other protocols make it a useful tool in designing, developing, analyzing, debugging and testing communication protocols that must be reliable and robust. As computing becomes more distributed over a computer network, applications will depend more on communication protocols that provide reliable ordered message delivery. The protocol development environment will enable protocol developers to meet the needs of these applications, now and in the future. ordered delivery protocol for interconnected local-area networks," In Proceedings 22 environment is to facilitate the development of reliable ordered message delivery protocols by providing a suitable and controllable environment in which to test and debug the target protocols. The simulation time of the development environment is a crucial component of accurate modeling of events in the system. Using multiple computers to perform distributed simulation introduces new problems in reeecting event occurrences. Approaches can be aggressive, simulating many activities concurrently on diierent processors but incurring a risk of rollback 15], or conservative where rollback is never required 24]. Lubachevsky's approach to event scheduling 21] is based on using what he calls the bounded lag time and minimum propagation delay to determine whether or not an event should be admitted for execution, thus reducing the chance of rollback. Our simulator uses a more conservative approach similar to window-based distributed simulation in which rollback cannot and does not occur. …

Cite this paper

@inproceedings{MelliarSmith2007theTS, title={\the Totem Single-ring Ordering and Membership Protocol," Acm Transactions on 6 Conclusion}, author={P. M. Melliar-Smith and Louise E. Moser and Deborah A. Agarwal and Ravi K. Budhia and P. Ciarfella and M. R. King}, year={2007} }