GraphLab Create on EC2 Cluster

User 113 | 3/24/2014, 4:32:13 PM

Hi,

Is it possible to deploy GraphLab Create on an EC2 cluster? I.e. Can SFrames store data across several machines or are they restricted to a single node? I'm essentially looking for functionality that's similar to GraphX's, where there are distributed graph and tabular data structures.

Thanks, Niko

PS: Any additional references concerning the technical details of SFrames would be appreciated. Have you published any white paper on the topic?

Comments

User 20 | 3/24/2014, 6:01:31 PM

Hi,

Scaling up both Graph and Table datastructures is an objective and is something we are working on. Right now we are focusing on single machine scaling and performance, but distributed is definitely on the horizon.

We have not yet published a white paper on SFrames. Since we are still making massive improvements to the SFrame architecture, any white paper now will be out of date really quickly :-). But putting out a white paper with technical details would be a good idea. We will look into it. (Maybe for the GraphLab conference?)

Yucheng


User 113 | 3/24/2014, 9:56:28 PM

Thanks, Yucheng. That's very useful to know.