Updating pagerank with iterative aggregation dating preparation image style

The timelines presented in the following were recorded using the graph.

Different machines’ timelines can differ significantly with Graph X, so we also supply the graphs for all machines.

Exchanging and sorting these data takes each worker just a few seconds (we use a radix sort), and it could be even faster with more effort.

We also profile both systems to explain the differences that we observe, address some of our readers’ questions from part one, and show some further improvements we have since made.

Finally, we return to our initial question of whether having a 10G network is beneficial.

It follows up a previous post evaluating the impact of fast networks on graph processing systems.

In the first part of this series, we have shown that some computations see large speedups when running on a faster network, illustrated by a Page Rank implementation in timely dataflow in Rust.

Search for updating pagerank with iterative aggregation:

updating pagerank with iterative aggregation-82updating pagerank with iterative aggregation-49updating pagerank with iterative aggregation-5

Leave a Reply

Your email address will not be published. Required fields are marked *

One thought on “updating pagerank with iterative aggregation”