Re: [jgrapht-users] Query about scalability using JGraphT

[Lokeya V. <lo...@gm...>]

Thanks a lot for your reply.

This gives me really a good idea. I like to how whats the size of a
node/vertex in the 500,000 vertex graph which you have mentioned.
My project is related to information processing to be more specific query
processing with the option for user feedback where they can take the initial
graph and further ask for more relations among documents, so the graph will
grow.

I understand the fact that the ability to get the graph is more dependent on
the RAM but then I just want to make sure if I have really a good amount of
memory support, then JGraph doesn't have any issue with respect to number of
nodes/edges.

On 4/10/07, Aaron Harnly <jgr...@li...> wrote:
>
> Hi Lokeya,
>
> On Apr 10, 2007, at 8:06 PM, Lokeya Venkatachalam wrote:
> > I will do that testing and see. But I am just curious if anyone has
> > done this before.
>
> I've done lots of work with about 500,000 vertices, which is a bit
> ungainly but certainly tractable.
>
> > To be more precise, will it scale to represent 1 million document ,
> > where each document is 1KB in size which is around 1GB of data in
> > form of nodes.
>
> Unless you're actively using the content of the documents as you
> traverse the nodes (certainly a possibility), it would seem wise to
> me to keep the node objects very lightweight, referring to a backing
> store of files or database rows as necessary. In the project I
> referred to above, we chose the latter route, with the nodes
> initially containing just id numbers, and optionally caching values
> from the database as needed.
>
> hope that helps,
> Aaron
> --
> Aaron Harnly
> Center for Computational Learning Systems
> Columbia University
>
>
>
>