PageRank

In its original context, the PageRank algorithm treated the web as a graph and was used to approximate the importance of a web page based on the number and importance of other webpages which link to it.

Lucata’s Page Rank implementation uses the traditional iterative approach in the “push” direction. It uses a remote atomic floating point add to push each vertex’s current score contribution to its neighbors, which is then used to compute the new scores at each vertex for the next iteration.

PageRank Benchmarks on Lucata

Scale1 Chassis4 Chassis
2030.7652.38
2242.83106.12
2450.36152.96
2649.89168.88
2848.40172.62
2948.73169.27
30172.31
* Data in MTEPS (millions of edges traversed per second)