MochiDB : A Byzantine Fault Tolerant Datastore
Published in CS244b: Distributed Systems, Autumn 2017, Stanford University , 2017
In this paper we would like to present MochiDB - a consistent, high volume, distributed datastore which is Byzantine fault tolerant. MochiDB supports native transactions and uses BFT quorum protocol with random write seeds that requires only two round trips for writes and one for read which gives it low latency over WAN deployments. This paper puts focus on engineering solutions that minimize the cost of contention resolution, sharding, dynamic configuration changes, garbage collection and others.
Recommended citation: Tsaturyan, T., & Dhakshinamurthy, S. MochiDB: A Byzantine Fault Tolerant Datastore https://www.scs.stanford.edu/17au-cs244b/labs/projects/tsaturyan_dhakshinamurthy.pdf