This is a simple Python simulation that can be used to test and experiment the Sumatra Simulation Management Tool.

Sumatra is a tool I found out about recently (10/12) that manages your computational experiments: simulations, calculations, analysis, any program your going to be running many times to create lots of data. It keeps track of what exactly you ran, in terms of source code, parameters, environment configuration, dependencies, etc.; when you ran it; how long it took; what was the input and output results. It lets you easily repeat configurations to be run again. And it let's you browse a library of your activity via your browser.

I've been looking for something like this ever since I ran my 1,000th simulation on our lab's computer cluster, I was so frustrated with keeping track of what I'm running. So I decided I'd put this tutorial online for the sake of others, and for demoing it to colleagues.

Note that demo this will not prevent you the need from installing and setting up Sumatra - but it will give you a quick simulation to run with it, and I provide some troubleshooting of problems I stumbled upon. If you find other problems and would like to add how you troubleshoot them, please fork this repo, edit the README.md file, and do a pull-request - I'd be delighted to merge. Note that this can be easily done in your browser using GitHub's text editor.

The simulation itself is pretty basic:

• Start off with a n individuals
• Divide them to two types, n/2 each
• In every simulation step, randomly sample the next generation so that each type gives birth to invdividuals of the same type
• Run x number of steps

n is given by a parameter called popsize and x by ticks. Both are at the params file under the default section.

The simulation is implemented in Python 2.7 and requires NumPy to be installed.

I did this on Windows, I did this on Linux. See the troubleshoot section for solving problems.

• Install Sumatra - pip install sumatra
• Install GitPython - pip install GitPython
• Get the code for this project by using git clone
• Setup Sumatra - smt init smt_test
• Configure Sumatra -
  • smt configure --executable=python
  • smt configure --main=main.py
• Run simulations: smt run params

Try different Sumatra stuff from here, like:

• smt run --label=some_label params to get a specific label for a record
• smt run --reason="why I ran this simulation" params
• smt run params default.popsize=100000 to run a simulation with more individuals
• smt list and smt list -l to see what already ran
• smt repeat some_label
• Try the web interface it's awesome: smtweb &

1. On Windows: when trying to call smt from the Windows command line it didn't work so I had to call python c:\python27\scripts\smt. I managed a workaround by working from Git Bash instead of cmd, but then found out a solution in the Sumatra Google Group.
2. On a Linux machine: I didn't have root control so installing and using Sumatra was a bit of tricky:

• First I installed [virtualenv] locally, that is: pip install --user virtualenv. If you are using a different python installer, check how to do it.
• Then I had to add the local site-packages folder to the PYHONPATH and the local bin folder to the PATH so that virtualenv would work. That took me some time because most tutorials assume bash but I'm using csh.
• Then I started a Sumatra virtual environment - see directions here. You got to be on bash to call bin/activate!
• Then I went on, and everything worked nicely.

1. On a Sun Grid Engine cluster, I had troubles running Sumatra because it uses git via GitPython which itself is not an implementation of git but a wrapper around the git executable (and a good wrapper it is!). So when GitPython uses a subprocess to call git I got Permission denied because I'm guessing SGE doesn't allow creating processes. Changing to an hg repository was the solution, because hg is implemented in python and therefore doesn't require the creation of a new process. If your remote repository must be in git (because, for example, you live in github), then you can still have the local repository in hg by using hg-git, an hg extensions that allows an hg client to talk to a git server. It's very easy to use so don't be intimidated. See the sge branch for an example of an .sge file suitable for running on a cluster.

The license for this repository (for Python and Sumatra license visit their website, but I think both are pretty much free):

This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.