GraphLab

Funding for the Next Generation of GraphLab

Joey — Tue, 14 May 2013 11:01:42 +0000

The GraphLab journey began with the desire:

to rethink the way we approach Machine Learning and Graph analytics,
to demonstrate that with the right abstractions and system design we can achieve unprecedented levels of performance, and
to build a community around large-scale graph computation.

We have been blown away by the excitement and growth of the GraphLab community and have been unable to keep up with the incredible interest from our amazing users.

Therefore, we are proud to announce GraphLab Inc, a company devoted to accelerating the development of the open-source GraphLab project.

Why a company?

Put simply, we need a full-time dedicated effort to take GraphLab from where it is today to where we would like to see it go in the future. With a dedicated team, you will see exciting new features, more integration with Cloud infrastructure, easier installation and deployment, along with a revamped support effort, with additional commercial-grade options and tools.

What happens to the open-source project?

The open source project will remain the flagship technology where we push the limits of graph computation and develop new ideas. With a dedicated team focused on the code, documentation, deployment process and most importantly support, you can expect a regular release schedule, faster turnaround on bug fixes and improvements, and a higher quality bar for the code going forward. We will continue to rely on our community to push us to tackle bigger, harder problems, and to contribute back to the open-source effort.

Are you going to charge for GraphLab now?

We believe the only way the best ideas in machine learning are developed is with the help of a vibrant community (it takes a village). That is why GraphLab will remain an open-source project; now with a company behind it, we can double our efforts to keeping the project healthy.

What is the roadmap for the open-source project?

GraphLab 2.2 is just around the corner, see here for more details as to what is in it. Beyond that, we are exploring a new computation engine and further enhancements to the communication layer, as well as simpler integration with existing Cloud technologies, easier installation procedures, and an exciting new graph storage system. And of course, we look forward to working with you to develop the roadmap and build the next generation of the GraphLab system.

The next 12 months will be an exciting time for us, and we hope that everyone will come along for the ride .

Learn more at graphlab.com

2nd GraphLab Workshop – Monday July 1st in SF

bickson — Mon, 13 May 2013 07:27:29 +0000

The 2nd GraphLab Workshop will take place on Monday July 1st at the Nikko Hotel, in downtown San Francisco.

Join us for a full day to learn about state-of-the-art development in Graph analytics solutions. Registration is now open – enjoy early bird discount rates!

The aim of the workshop is to bring together researchers from academia, as well as data scientists from industry with the special focus of large scale machine learning on sparse graphs.

We have secured talks and demos about the hottest graph processing systems and graph storage out there:

Google (Pregel)
GraphLab (CMU/UW)
Giraph (Facebook)
Cassovary (Twitter)
Naiad (Microsoft Research)
GraphBuilder (Intel Labs)
Presto (HP Labs)
Grappa (UW)
Combinatorial BLAS (LBNL/UCSB)
Allegro Graph (Franz)
Neo4j (Neo Technology),
Titan (Aurelius)
DEX (Sparsity Technologies)
YarcData

Also learn about the latest developments and future directions of the GraphLab project.

We look forward to seeing you in July!

Preliminary Agenda

Program Committee

Our sponsors

An update: we just learned that each participant will get a free copy of the book Graph Databases – thanks to Neo Technology for sponsoring!
- GraphLab Team

Exciting Updates on the Next Release

Joey — Mon, 29 Apr 2013 21:06:10 +0000

So it’s been a while …

We know, but we’ve been busy! We are putting the finishing touches on GraphLab 2.2 and expect to release it in early July at the GraphLab workshop.

What you can expect in this next release of GraphLab is:

Substantially improved loading and runtime performance through new partitioning and layout methods
Nearly an order of magnitude improvements in asynchronous engine performance.
Substantial improvements in the network communication layer.
New additional simplified vertex-programing API using the WarpEngine.
New toolkits functionality including dual decomposition.
Improvements in code quality, testing, and documentation.

Did we mention GraphLab 2.2 is faster?

Early benchmarks suggest factor of 2 improvement in PageRank.
Triangle Counting Runtime for the Twitter Graph (used to take 90s on 64 cc2.8xl machines), is now 15s using the same configuration. That is a 6x speedup.

Oh, and one more thing… we are forming a company to take the GraphLab project to the next level.

CLiMF algorithm is now implemented in GraphChi

bickson — Thu, 18 Apr 2013 14:40:24 +0000

Many thanks to Mark Levy (last.fm) for contributing implementation of the CLiMF algorithm to our growing collaborative filtering toolkit: read more.

Distributed Dual Decomposition is now implemented in GraphLab

bickson — Thu, 18 Apr 2013 14:37:47 +0000

Many thanks to Dhruv Batra from Virginia Tech for kindly donating this implementation to our growing
graphical models toolkit. Full documentation is here.

PowerGraph Presented at OSDI

Joey — Mon, 08 Oct 2012 23:42:09 +0000

We just presented our work on PowerGraph at OSDI! PowerGraph is the system and abstraction that forms the foundation of the current version of GraphLab. The PowerGraph research effort attempts to answer the question:

“How can we efficiently design and implement distributed computation on the large-scale natural-graphs graphs found in settings ranging from social networks to the web?”

Checkout the slides from the talk as well as our paper.

Two OSDI Papers

ylow — Sun, 09 Sep 2012 19:51:15 +0000

We have two papers accepted to OSDI 2012!

One on GraphLab:

PowerGraph: Distributed Graph-Parallel Computation on Natural Graphs. Joseph Gonzalez, Yucheng Low, Haijie Gu, Danny Bickson and Carlos Guestrin. [pdf][bib/abs]

And another on GraphChi:

GraphChi: Large-Scale Graph Computation on Just a PC. Aapo Kyrola, Guy Blelloch, and Carlos Guestrin. [pdf] [bib]

GraphLab Keynote

Joey — Tue, 10 Jul 2012 17:52:34 +0000

Carlos introduced the next generation of the GraphLab abstraction including the exciting new GraphChi project. Checkout the slides:

Download Keynote Slides

GraphLab Workshop 2012 is a Supernode!

Joey — Tue, 10 Jul 2012 17:29:54 +0000

We want to thank everyone who attended the GraphLab 2012 workshop on BigLearning. It was a great success! We had over 300 attendees and a density of exciting talks that rivals top tier conferences. For those who missed it, we will be posting the slides and posters as well as the video lectures as soon

Intel GraphBuilder

Joey — Mon, 09 Jul 2012 13:17:18 +0000

We have been collaborating with Intel to develop the new GraphBuilder library which provides tools to construct large-scale graphs on top of Apache Hadoop. The library provides function for:

Pre-processing – Feature selection/Tokenization, and Tabulation
Graph construction – Edge and Vertex lists
Graph Normalization – Compression techniques for sparse graph labels
Graph Transformation – Optional filters for self- and duplicate edge detection, as well as directed to undirected transformation
Graph Partitioning – Multiple schemes to partition graphs for GraphLab v2.1 ingress
Serialization – Supports JSON serialization

Open source library coming soon!