Pregel: Large-scale graph computing at Google

googleresearch.blogspot.com

36 points·jashmenn··10 comments

Open articleView on HN

Does anyone have any more information about this?

Is it built on top of MapReduce or is it totally separate from the ground up?

And is the paper from last year available anywhere? Are examples of implementations available?

Apprently this is the paper: http://projects.will.madstones.com/thesis/papers/Malewicz%20...

But it's just 1 page, a bit of a letdown...

http://portal.acm.org/citation.cfm?id=1582716.1582723

only paper I found. It's only a page though!

But it's on the accepted papers list for sigmod/pods 2010, so with luck in a month or so we'll know more. http://j.mp/aPJuyA

I just checked with my coworker Marko (author of Gremlin [1]) and he told me that no-one really has much info on this. Though he said that Apache Hama [2] + Angrapa [3] is a supposed to be a clone.

I think Hama is built on MR so maybe Google's version is as well?

[1] http://github.com/tinkerpop/gremlin

[2] http://incubator.apache.org/hama/

[3] http://wiki.apache.org/hama/GraphPackage

http://www.royans.net/arch/pregel-googles-other-data-process...

http://blog.udanax.org/2009/08/inference-anatomy-of-google-p...

http://portal.acm.org/citation.cfm?id=1073981

also check out PIs publications and his thesis http://research.google.com/pubs/author145.html

The lead author will also be giving a presentation on Pregel at SIGMOD 2010: http://www.sigmod2010.org/research_list.shtml

"Implementing PageRank, for example, takes only about 15 lines of code."

Interesting. I wonder what it looks like.

It's a white powdery substance... oh, code, never mind.

I wonder if this can be considered an NP-complete oracle.

Reply on news.ycombinator.com