PnyxDB: a Lightweight Leaderless Democratic Byzantine Fault Tolerant Replicated Datastore

TL;DR

PnyxDB is a scalable, low-latency Byzantine Fault Tolerant datastore that uses conditional endorsements and supports application-level voting, enabling efficient governance for large, geodistributed systems.

Contribution

It introduces conditional endorsements and application-level voting to improve scalability and flexibility in Byzantine Fault Tolerant datastores.

Findings

Speeds up commit latencies by a factor of 11

Maintains latency below 5 seconds in 180-node deployment

Outperforms BFTSMaRt and Tendermint in benchmarks

Abstract

Byzantine-Fault-Tolerant (BFT) systems are rapidly emerging as a viable technology for production-grade systems, notably in closed consortia deployments for nancial and supply-chain applications. Unfortunately, most algorithms proposed so far to coordinate these systems suffer from substantial scalability issues, and lack important features to implement Internet-scale governance mechanisms. In this paper, we observe that many application workloads offer little concurrency, and propose PnyxDB, an eventually-consistent Byzantine Fault Tolerant replicated datastore that exhibits both high scalability and low latency. Our approach is based on conditional endorsements, that allow nodes to specify the set of transactions that must not be committed for the endorsement to be valid. In addition to its high scalability, PnyxDB supports application-level voting, i.e. individual nodes are able to endorse or reject a transaction according to application-defined policies without compromising consistency. We provide a comparison against BFTSMaRt and Tendermint, two competitors with different design aims, and show that our implementation speeds up commit latencies by a factor of 11, remaining below 5 seconds in a worldwide geodistributed deployment of 180 nodes.