[Release][Blog] Accelerating Partitioning of Billion-scale Graphs with DGL v0.9.1

Graphs is ubiquitous to represent relational data, and many real-world applications such as recommendation and fraud detection involve learning from massive graphs. As such, GNNs has emerged as a powerful family of models to learn their representations. However, training GNNs on massive graphs is challenging, one of the issues is high resource demand to distribute graph data to a cluster. For example, partitioning a random graph of 1 billion nodes and 5 billion edges into 8 partitions requires a powerful AWS EC2 x1e.32xlarge instance (128 vCPU, 3.9TB RAM) running for 10 hours to finish the job.

This is a companion discussion topic for the original entry at https://www.dgl.ai/release/2022/09/19/release.html