Experiment scripts

In this repository we most of the scripts used to reproduce the experiments from the paper "Analysis of Scalability in GraphChi".

The packages needed to generate the graphs and run the scripts are:

• snap-stanford 6.0.0
• numpy (any recent version).

Moreover, one needs to have a local copy of the GrahphChi software.

Quick description of the scripts:

• cc_interface and avgtimeexp_proxy.py were used to perform the experiments on compute canada and do not have much purpose out of compute canada servers;
• erdos_renyi.py generates graphs with edges uniformly distributed;
• gen_cycle.py generates 3-regular circular graphs as described in the paper;
• grid_generator.py generates grid graphs as described in the paper;
• rmat_gradient.py generates a range of different RMAT graphs, going from real-world like to uniformly distributed edges;
• main_experiments.py this files performs the experiments as described in the paper, with warmup round and varying number of threads. The list of files the script uses is hardcoded. Also, note that the path information for GraphChi at the beginning of the script need to eb updated for the program to be able to find the GraphChi executables. It assumes that the example apps from GraphChi are already compiled.

PS: The information on this README are not great at this point, but I plan to add a bit more of information soon. However, I don't know if this would be fair given that these updates would happen after the official deadline, so that is why I uploaded a rough draft of the instructions before Apr 30 5pm.