Use cuGraph to Improve Performance of contribution and collaboration Services #58
Description
Description
The contribution and collaboration services, despite using graph-tool, are not a performant as we would like to collect the respective metrics from a large project like MySQL.
The initial approach to improving performance was to use PageRank centrality instead of node betweenness centrality in the contribution service (See Issue #53). A preliminary exploration of this approach did not yield marked improvements in performance. Furthermore, while PageRank centrality is likely to be correlated with node betweenness centrality, there is no equivalent for the edge betweenness centrality used by the collaboration service. Fortunately, the computation of betweenness centrality measures from a graph is a problem that can benefit from computation in a GPU and cuGraph is a GPU Graph Analysis library that is shown to significantly improve performance of betweenness centrality computation.