Share |

EMI services used in neuro-scientific grand challenge

Name: 
EMI Project Office
Organization: 
CERN

EMI [1] teamed-up with a number of projects in a large-scale trans-continental scientific computing challenge. LINGA (LInked Neuroscientific Grand chAllenge) is the first data and compute challenge involving neuro-scientific infrastructures from two continents – neuGRID [2] (Europe), CBRAIN [3] (North America) and LONI [4] (North America), and distributed computing infrastructures – EGI’s Life Science Computing Grid  [5] (LSCG, Europe) and EMI resources [6] (Europe). Some 11,000 brain scans from distributed neuro-centers were analyzed using computationally demanding image processing applications for marker discovery in neuro-degenerative diseases like Alzheimer’s.  

The challenge is not only due to the scale and distributed nature of the data sets or the complexity of the image processing applications in use; the participating infrastructures use different underlying compute management technologies. SHIWA [7] and outGRID [8] projects developed a framework that orchestrates these heterogeneous infrastructures such that their underlying compute solutions are harnessed to process brain scans coming from different sources and handle different ‘stages’ of the image processing applications. Processing is done in parallel and asynchronously, with the aggregated output further processed by LSGC. Two important components are extracted to explain the outputs’ variability and potential correlations between individuals’ characteristics and the disease development. If this activity was run in a single neuro-scientific centre, it would have taken 50 days. Through this collaboration, the challenge concluded in seven days using 3,000 CPUs per processing cycle. The LINGA challenge was presented at the EGI Technical Forum 2011 [10] where it won the Best Demonstration competition.
 
For this scientific challenge, EMI has made available its demonstration testbed with the latest services from EMI 1 Kebnekaise. Specifically the EMI services EMI-BDII (top and site), EMI-CREAM, EMI-WMS and EMI-WN provided one set of the compute technologies leveraged by the SHIWA-outGRID framework. This collaboration has been an excellent opportunity for EMI to showcase its first release, EMI 1 Kebnekaise, and is a good demonstration of the stability and usability of EMI 1 services for large-scale scientific research.  

For more information, please contact:
EMI: EMI Project Office (emi-po@cern.ch)
LINGA challenge: Giovanni Frisoni (giovanni.frisoni@gmail.com), David Manset (dmanset@maatg.fr)

[1] EMI: http://www.eu-emi.eu
[2] neuGRID: http://www.neugrid.eu
[3] CBRAIN: http://cbrain.mcgill.ca
[4] LONI: http://www.loni.ucla.edu   
[5] LSCG: http://www.egi.eu/collaboration/LSGC.html
[6] EMI Demonstration Testbed, Section 4.5: https://twiki.cern.ch/twiki/bin/view/EMI/TestBed#4_Testbed_Resources_Des...
[7] SHIWA: http://www.shiwa-workflow.eu
[8] outGRID: http://www.outgrid.eu
[9] EMI 1 Kebnekaise: http://www.eu-emi.eu/emi-1-kebnekaise