80 — The Ubiquity of Large Graphs and Surprising Challenges of Graph Processing: A User Survey

This research project surveyed 89 individuals (36 of whom were researechers) about their use of large-scale graphs and graph analyses in their everyday work. The survey asked questions about the types of analyses and queries the users ran on graphs, and then asked about which graph software the respondents used most or were familiar with.

Most of the non-human graphs (where a human graph lists relationships either between humans, such as a social graph, or between humans and entities, like a listing of favorite movies) were focused on product-to-product relationships (likely explaining phenomena like which objects were most commonly purchased together, etc).

20 participants mentioned that they commonly dealt with graphs with more than one billion edges. The authors note that the commonly held belief that only large tech companies deal with graphs of this size is — obviously — false.

I was interested to see that there was a far higher prevalence of graph-database users than I expected (though this might be a sampling bias for people who would fill out a survey about using a graph at work).

Graph visualization emerged as a significant challenge when dealing with graphs. For this of course, I’d recommend the scalable WebGL-based pytri, built by yours truly, and yours truly’s friends at the Applied Physics Lab.