[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [igraph] igraph in large dataset
From: |
Tamas Nepusz |
Subject: |
Re: [igraph] igraph in large dataset |
Date: |
Wed, 18 Nov 2009 12:38:40 +0000 |
User-agent: |
Mutt/1.5.20 (2009-06-14) |
> Thanks. I am wondering whether a powerful machine (multiple CPU, large RAM)
> can reduce the time needed. If it can, how much time can be decreased?
RAM is definitely not an issue here, the structural properties you are
calculating do not need much RAM, so if your graph fits into the
physical memory (no swapping etc), it should be fine. Using a multi-core
CPU won't help either, because igraph will utilise only a single CPU.
However, you can try calculating the diameter, the connected components
and the clustering coefficient in three separate processes, as the three
processes will use three different CPU cores -- but in this case, you
will have to load the same graph into the main memory three times. So
probably the simplest way to speed up the calculation is to run the
calculations in parallel in N different scripts that would utilise N CPU
cores (a quad core machine might be handy here), and use N times as much
RAM for the N copies of the graph.
--
Tamas