Simulations and data analyses within TVB project are performed on our High Perfomance Computing plateform. The exploration of high dimensional parameters space is achieved in a reasonable time using distributed computing techniques and grid computing engines. It allows to manipulate ranges of network parameters and identify quickly critical sets of parameters regarding the whole brain dynamic.
This plateform is an heterogeneous HPC, accessed directly by TVB scientist community or remotely via a web-based interface: tvb interface.
Hardware is divided in two parts :
– Virtual applications servers which allows basic services such as login, sys admin, data basing, web … .
– Computing servers: the main part comprises grid computing based nodes for a total of 240 computing cores (Intel Xeon E5-2670 CPU). In addition, there is a GPU computing node composed of 4 Nvidia Tesla M2075 GPUs for a total of 1792 cores, allowing to run fast CUDA based applications.
Jean-Luc Blanc, technical director
Loic Bonnier, external collaboration
New users please check *Cluster usage policy*