Tornando Paxos Mais Escalável com Réplicas Leitoras

• Anderson de Paula UFSCar
• Gustavo Vieira UFSCar

Abstract

Paxos is an active replication algorithm that keeps the same shared state consistently among servers that handle requests from a distributed application. In this paper we propose a reconfiguration mechanism for the Paxos algorithm that allows the incorporation of new replicas in the system without increasing significantly the cost to keep the whole system consistent. Our approach uses reader replicas, capable to answer all application requests without taking an active part in the costly operations of the Paxos algorithm. We were able to add new servers to a computation in such a way that a high load could be serviced, without compromising the application throughput.

References

Lamport, L. (2006). Fast Paxos. Distrib. Comput., 19(2):79–103.

Lamport, L., Malkhi, D., and Zhou, L. (2010). Reconguring a state machine. SIGACT News, 41(1):63–73.

Schneider, F. B. (1990). Implementing fault-tolerant services using the state machine approach: a tutorial. ACM Comput. Surv., 22(4):299–319.

Vieira, G. M. D. and Buzato, L. E. (2008). Treplica: Ubiquitous replication. In SBRC '08: Proc. of the 26th Brazilian Symposium on Computer Networks and Distributed Systems, Rio de Janeiro, Brasil.

Vieira, G. M. D. and Buzato, L. E. (2009). The performance of Paxos and Fast Paxos. In SBRC '09: Proc. of the 27th Brazilian Symposium on Computer Networks and Distri- buted Systems, pages 291–304, Recife, Brasil.

Vieira, G. M. D. and Buzato, L. E. (2010). Implementation of an object-oriented speci- cation for active replication using consensus. Technical Report IC-10-26, Institute of Computing, University of Campinas.