GraphLab Crate Canvas scalability?

User 5158 | 6/1/2016, 2:26:07 AM

Hi, I am interested to use GraphLab Crate Canvas for researching large graph visualisation on distributed Amazon machines. I am concerned with big data and would like to know what sized graphs can GraphLab handle? If there are any papers available I would greatly appreciate it. Side note: currently experimenting with Apache Giraph and would like to see how GraphLab Crate Canvas works on big data.

Comments

User 2014 | 6/1/2016, 9:02:45 PM

Hi,

Graphlab Create can handle large amounts of data, through the scalable out-of-core data structure that are the SFrame and SGraph. Canvas is our visualization product that allows you to peek at your data and plot different things.

Could you elaborate further on what you are looking for? There is no paper per-se on Canvas, but you can find the documentation at: https://dato.com/products/create/docs/graphlab.canvas.html

  • Antoine

User 5158 | 6/2/2016, 2:09:05 AM

Thanks for your reply Antonie, I am testing different frameworks and performing some research on them. I would like to draw graphs of different sizes from 1,000 to 1,000,000+ edges if possible. I will be running GraphLab Canvas on multiple Amazon EC2 machines. Thanks.


User 2014 | 6/2/2016, 7:47:42 PM

Hi,

For computation, our system can totally handle it with no problem. For visualization, it depends how much information you want to display. A large graph will look cluttered on a plot if there are too many things to display.

So, display-wise, I can't guarantee that it will look good ;) But computation-wise, our platform can handle it.

Antoine