Abstract
This paper describes and evaluates an atomic multicast protocol that guarantees order and agreement conditions. The latency of message delivery in the new protocol is lower latency than previous protocols that provide the same functionality. The new protocol achieves low latency by eagerly delivering the multicast message to the application program before reaching agreement. on the receipt order. Thus, the application can process the message and proceed in parallel with the execution of the agreement protocol, instead of blocking until all receivers reach consensus as in previous protocols. Results from a prototype implementation show that by using the proposed protocol instead of a traditional one, the latency of common file-system remote procedure calls can be reduced by up to 23%.
Original language | English (US) |
---|---|
Title of host publication | WORLD MULTICONFERENCE ON SYSTEMICS, CYBERNETICS AND INFORMATICS, VOL 1, PROCEEDINGS |
Publisher | INT INST INFORMATICS & SYSTEMICS |
Pages | 541-547 |
Number of pages | 7 |
ISBN (Print) | 980-07-5912-3 |
State | Published - 1999 |
Externally published | Yes |
Keywords
- distributed systems
- fault tolerance
- multicast protocols
- replication
- COMMUNICATION