Project

Realizing the Full Potential of RDMA Technology in Accelerating Distributed Consensus

Principal Investigator

School of Computer Science, University of Waterloo, Canada

Summary

Remote Direct Memory Access (RDMA) provides extremely high networking performance that can radically affect foundational distributed systems algorithms.  In fact, approaches have already been developed to enhance performance of popular distributed consensus protocols such as Paxos and Raft.  We however argue that these works do not fully leverage the parallelism enabled by direct remote access.  This project aims to exploit all the capabilities provided by the rich RDMA interface to deliver a new generation of radically different distributed consensus protocols.