Graphing directed graphs with lots of nodes and vertices

Nonius · Post by **Nonius** » Sat Jan 28, 2012 5:56 pm

Been playing around with it.

graphed Gnutella network data from August of 2008.

Gnutella peer-to-peer network, August 8 2002

Dataset information

A sequence of snapshots of the Gnutella peer-to-peer file sharing network from August 2002. There are total of 9 snapshots of Gnutella network collected in August 2002. Nodes represent hosts in the Gnutella network topology and edges represent connections between the Gnutella hosts.

Dataset statistics

Nodes 6301

Edges 20777

Nodes in largest WCC 6299 (1.000)

Edges in largest WCC 20776 (1.000)

Nodes in largest SCC 2068 (0.328)

Edges in largest SCC 9313 (0.448)

Average clustering coefficient 0.0150

Number of triangles 2383

Fraction of closed triangles 0.02066

Diameter (longest shortest path) 9

90-percentile effective diameter 5.5

Source (citation)

J. Leskovec, J. Kleinberg and C. Faloutsos. Graph Evolution: Densification and Shrinking Diameters. ACM Transactions on Knowledge Discovery from Data (ACM TKDD), 1(1), 2007.

M. Ripeanu and I. Foster and A. Iamnitchi. Mapping the Gnutella Network: Properties of Large-Scale Peer-to-Peer Systems and Implications for System Design. IEEE Internet Computing Journal, 2002.

from Stanford's snap

GnuNonius

tristanreid · Post by **tristanreid** » Wed Oct 31, 2012 6:11 pm

Hey Nonius, this thread came out during my hiatus from the NP-Land-of-Interestingness.

I've been using a python library called pygraph (implements the graph data structures and algorithms) plus pydot (implements output to DOT language, uses GraphViz to make pretty pictures).

I'm looking at the Gephi site - it looks amazing, and much more polished than the stuff I've been using. Sadly, I can't access youtube from work, so I can't look at GnuNonius until I get home tonight (and post-Halloween festivities). I assume you found a way to make video?

So far I'm using this for some pretty simple stuff, mostly IT-related. I think I've found some pretty generic problems in "Enterprise IT" that are clear fits for graph-analysis, but nobody seems to be using them, enough so that I actually thought about doing a start-up based around what I've made so far (I'm not willing/capable to do the sales part - non-starter). The only algorithms I'm using are bfs/dfs, finding cycles, spanning trees, and some rule-based coloring that helps to find critical nodes. I've gotten a couple of automated-wins out of the algorithms, but the main reason people like it is the pretty graphs. That's why I'm wondering about taking it to the next level, prettyness-wise. Any tips on the goodness or badness of Gephi?

-t.

Nonius · Post by **Nonius** » Fri Nov 02, 2012 11:17 pm

Hi Tristanreid, no, I didn't make the video the way I wanted to. What I did was create a random graph and applied one of their off the shelf algos to configure the graph "optimally". That process is animated on the screen, so I did a screen recording. Gephi is pretty cool, but it is a bit slow and what I'd like is a) ability to make a video of a dynamic graph according to dynamics I specify and b) a way to implement homegrown algos to configure the graph.