A Comprehensive Dynamic-Graph System for Modern Applications

A unified system for storing, updating, and analyzing large, skewed, dynamic graphs.

Real-world graphs are large, skewed, and constantly changing. This project builds a comprehensive system for dynamic graphs — efficient containers, a framework that supports updates and analytics together, and the benchmarks needed to evaluate them.

We introduced Terrace (Pandey et al., 2021), a hierarchical container for skewed dynamic graphs, and BYO (Wheatman et al., 2024), a unified framework for benchmarking large-scale graph containers. We have also optimized communication for the distributed execution of graph neural networks (Kurt et al., 2023).

References

2024

  1. VLDB 2024
    BYO: A Unified Framework for Benchmarking Large-Scale Graph Containers
    Brian Wheatman, Xiaojun Dong, Zheqi Shen, and 4 more authors
    Proc. VLDB Endow., 2024

2023

  1. IPDPS 23
    Communication Optimization for Distributed Execution of Graph Neural Networks
    Süreyya Emre Kurt, Jinghua Yan, Aravind Sukumaran-Rajam, and 2 more authors
    In IEEE International Parallel and Distributed Processing Symposium, IPDPS 2023, St. Petersburg, FL, USA, May 15-19, 2023, 2023

2021

  1. SIGMOD 2021
    Terrace: A Hierarchical Graph Container for Skewed Dynamic Graphs
    Prashant Pandey, Brian Wheatman, Helen Xu, and 1 more author
    In SIGMOD ’21: International Conference on Management of Data, Virtual Event, China, June 20-25, 2021, 2021