An exploration of the role of synchrony on the performance of Paxos

• Grant number: 13/21651-0
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate (Direct)
• Start date: May 01, 2014
• End date: April 30, 2015
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Luiz Eduardo Buzato
• Grantee: Daniel Cason
• Supervisor: Fernando Pedone
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Institution abroad: Università della Svizzera italiana, Switzerland
• Associated to the scholarship: 11/23705-4 - Total order broadcast: an exploration of the role of synchrony on performance of Paxos, BP.DD

Abstract

Total order broadcast and consensus are at the core of practical solutions for the problem of creating highly available distributed applications based on active replication. More specifically, in asynchronous computation systems which processes can fail these two problems are equivalent: we use consensus instances to determine the ordered sequence of messages that should be delivered to the processes. Paxos is an optimal consensus algorithm devised for the asynchronous crash-recovery computing model augmented with unreliable failure detectors. By using it to generate an ordering for the broadcast messages, Paxos guarantees that no replica state is going to diverge from each other despite the occurrence of failures and the asynchronicity of processes and communication channels. In this project, we are interested in exploring the behaviour of Paxos within the alternative framework provided by timed-asynchronous computing model, augmented by a simple pulse-based mechanism intended to synchronize the processes' execution. In this context, our objective is to verify whether it is possible to devise a variation of Paxos to this computation environment, that takes advantage of its synchrony to achieve a better performance, but still providing the same guarantees of the original algorithm. (AU)