Monday, March 3, 2014

Docker image to run the software neural simulators used in BrainScaleS

Docker is like a virtual machine (Virtualbox, VMware,...) but with one huge advantage for everyone wanting to run a neural simulator on his own computer: being based on Linux containers, a Docker image can use all the resources (CPUs, RAMs) of your host computer.
That's why we have decided to put all the tools needed to run the software neural simulators used in the European project BrainScaleS into an openly accessible docker image. To use it, please follow the official step-by-step guide.
This Docker image (brainscales/neural-networks:software) contains the following neural simulators and the associated tools (openmpi, visualization,...):
  • NEST (version 2.2.2)
  • NEURON (version 7.3)
  • Brian (latest version from PyPI, currently 1.4.1)
  • PyNN (latest version from PyPI, currently 0.7.5), the language-independent simulator (it uses as backend either NEST, Brian or NEURON),
  • MUSIC, the multi-simulation coordinator (multiconn branch). Note that, in this image, only NEURON has been compiled with MUSIC.
To use this image, you need to install Docker on your host computer. On an Ubuntu computer, this means installing lxc-docker:
  $ sudo apt-get install lxc-docker
Then, you download the image from the Docker index repository:
  $ sudo docker pull brainscales/neural-networks:software
You can now follow this step-by-step guide to run the examples from all these neural simulators using all the resources from your host computer, without having to install other software than Docker itself and a ssh client. Enjoy!

Joël Chavas and Andrew Davison.