Performance Comparison Between the Paxos and Chandra-Toueg Consensus Algorithms
Naohiro Hayashibara, André Schiper, Péter Urbán, Takuya Katayama, Ecole Polytechnique, Fédérale De Lausanne · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 2002
Protocols which solve agreement problems are essential building blocks for fault tolerant distributed applications.While many protocols have been published, little has been done to analyze their performance.This paper represents a starting point for such studies, by focusing on the consensus problem, a problem related to most other agreement problems.The paper compares the latency of two consensus algorithms designed for the asynchronous model with failure detectors: the Paxos algorithm and the Chandra-Toueg algorithm.We varied the number of processes which take part in the execution.Moreover, we evaluated the latency in different classes of runs: (1) runs with no failures nor failure suspicions, (2) runs with failures but no wrong suspicions.We determined the latency by measurements on a cluster of PCs interconnected with a 100 Mbps Ethernet network.We found that the Paxos algorithm is more efficient than the Chandra-Toueg algorithm when the process that coordinates the first round of the protocol crashes.The two algorithms have almost the same performance in all other cases.