igraph-help
[Top][All Lists]

## Re: [igraph] generating large graphs

 From: Gábor Csárdi Subject: Re: [igraph] generating large graphs Date: Sat, 27 Jul 2013 13:16:15 -0400

According to the formula in my previous mail this requires about 108GB of storage, plus some extra (~50GB, maybe more) when creating the graph.

Maybe your operating system does not support allocating huge chunks of memory. E.g. can you create an array of length 200 * 2**25 (number of edges times 2)? This is needed to create the graph.

Gabor

On Sat, Jul 27, 2013 at 10:33 AM, Claudio Martella wrote:
Hi Gabor,

here it is.

#!/usr/bin/python

import igraph

vertices = 2**25
edges = 100
g = igraph.Graph.Barabasi(vertices, edges)
g.write("%s-graph.txt" % vertices, format="edgelist")

On Sat, Jul 27, 2013 at 3:21 PM, Gábor Csárdi wrote:
The memory you need for a graph is (2*n + 4*m) * 8 bytes, where 'n' is the number of vertices and 'm' is the number of edges. So if the density of the graph is not too large, this should be possible. Can you send some code that reproduces this?

Gabor

On Sat, Jul 27, 2013 at 8:37 AM, Claudio Martella wrote:
Hello guys,

I've been trying to build a graph with over 32M vertices using Barabasi() via the python wrapper. The machine I'm currently using has around 160G of main memory but the process is being OOM-killed by the kernel. I was able to generate a 16M vertices graph with a usage of around 50% of memory. Do you guys have some suggestions about how i can generate such large graphs (potentially in a distributed fasion), as I'm actually aiming also at larger ones?

Thanks,
Claudio

--
Claudio Martella

_______________________________________________
igraph-help mailing list
https://lists.nongnu.org/mailman/listinfo/igraph-help

_______________________________________________
igraph-help mailing list
https://lists.nongnu.org/mailman/listinfo/igraph-help

--
Claudio Martella