Understanding fault-tolerant distributed systems
Communications of the ACM
Replication management using the state-machine approach
Distributed systems (2nd Ed.)
Distributed systems (2nd Ed.)
Hi-index | 0.00 |
To the wide area network oriented distributed computing, a slow service is equivalent to an unavailable service. It makes demands of effectively improving the performance of the replication algorithms without damaging the availability. Aim for improving the performance of the active replication algorithm, we propose a new replication algorithm named RRR (Rapid Response Replication). Its basic idea is: all replicas receive request, but only the fastest one sends back the response to the client after it handles the request. Its main advantages are: (1) In the algorithm, the response is sent back directly by the fastest replica after it handles the request; (2) The algorithm avoids agreement; (3) The algorithm avoids the redundant nested invocation problem (the problem may be arisen while using active replication).