Part of my post on cuGraph and Rapids.
In my benchmark study, I use three datasets in increasing order of scale from the Stanford Large Network Dataset Collection.
- ego-Facebook: Undirected graph with 4 K nodes and 88 K edges from Facebook
- ego-Twitter: Directed graph with 81 K nodes and 1.7 M edges from Twitter
- ego-Gplus: Directed graph with 107 K nodes and 13.6 M edges from Google+