Re: [igraph] benchmarks?

[Tamas Nepusz]

Dear Michael,

> Does anyone have any rough benchmark numbers for iGraph running on a
> sparse network of 100k nodes and 500k edges or lager? Ideally I'm
> looking for a benchmark for eigen value centrality score, but any basic
> metric would do.
I did a quick test using the Python interface (which has roughly the
same overhead as the R interface, so the results should be the same) and
using PageRank instead of eigenvector centrality some time ago. I
used a network of 2.13M vertices and 7.56M directed edges:

>>> g=load("test_graph.lgl", directed=True)
>>> print g
Directed graph (|V| = 2137884, |E| = 7560862)
>>> from time import time
>>> t1=time(); ec=g.pagerank(); t2=time(); print t2-t1
44.2914

(The time above is in seconds).

On a smaller graph with 64K vertices and 240K edges, the PageRank
calculation (which is very similar to eigenvector centrality) takes
about 0.5s.

These results were achieved on a single core of an Intel Xeon X3360
running at 2.83 GHz.

-- 
Tamas