Distributed Consensus Made Fast

Speaker: Gengrui (Edward) Zhang, PhD
Assistant Professor
Department of Electrical and Computer Engineering
Concordia University

Title: Distributed Consensus Made Fast

Date: Thursday, 24 July 2025

Time: 2:00pm - 3:00pm

Venue: Room 2463 (via lift 25/26), HKUST

Abstract:

Consensus algorithms are foundational to modern distributed applications—powering everything from database transactions and blockchain protocols to cloud coordination and distributed machine learning. As system scales grow and heterogeneity becomes the norm, traditional protocols like Paxos and Raft, built on static majority quorums, struggle to keep up with performance demands. This talk introduces a new family of weighted consensus algorithms that dynamically assign node weights to accelerate agreement and enhance system efficiency. We highlight Cabinet, a novel protocol that significantly outperforms Raft in large-scale and heterogeneous environments. Through extensive evaluation on MongoDB and PostgreSQL using YCSB and TPC-C workloads, Cabinet demonstrates superior throughput and lower latency across increasing system sizes, complex network conditions, and failure scenarios—across both homogeneous and heterogeneous clusters. We conclude with broader insights into consensus system design and discuss emerging directions for building high-performance, resilient distributed systems.


Biography:

Dr. Gengrui (Edward) Zhang is an Assistant Professor in the Department of Electrical and Computer Engineering at Concordia University, where he leads the Distributed Computing and Systems (DISCOS) Research Group. Dr. Zhang’s research focuses on advancing the development of high-performance, scalable, and reliable distributed systems. His work addresses real-world challenges in key areas of AI-supporting distributed systems, blockchain, cloud computing, and data management.

Dr. Zhang obtained his PhD from the University of Toronto in 2024. He has served as program committee member of various international conferences, including IEEE DAPPS, and ACM SYSTOR, ACM Middleware, IEEE ICDE, and VLDB. He actively collaborates with both academia and industry.