Hao
Tan,
Master’s
candidate
David
R.
Cheriton
School
of
Computer
Science
All-to-all data transmission is a typical data transmission pattern in both consensus protocols and blockchain systems. Developing an optimization scheme that provides high throughput and low latency data transmission can significantly benefit the performance of those systems.
In this talk, I will present ShallowForest, a data plane optimization for consensus protocols deployed in wide-area networks (WAN). The goal of ShallowForest is to improve consensus protocols' resilience to skewed client load distribution. ShallowForest can dynamically adjust data transmission overlays according to client load and available bandwidth at each replica to make consensus protocols achieve high throughput and low latency. Experiments with skewed client load across replicas in the Amazon cloud demonstrate that ShallowForest can improve the commit throughput of the EPaxos consensus protocol by up to 100% with up to 60% reduction in commit latency.