Mapping out needs and requirements for GPU based computing in the European Grid
The use of GPGPUs or co-processor accelerator devices, such as Intel's Xeon Phi co-processor, over the past few years has resulted in significant increases in the performance of certain applications. With three of the top 10 clusters in the current June 2012 Top500 supercomputers list using NVIDIA GPGPUs, we would expect the number of GPGPU or co-processor deployments at grid resource centres to grow significantly over the next few years.
GPU based computing is definitely gaining momentum in the scientific worlds, so the EGI-InSPIRE project setup a 'Virtual Team' to map out needs and requirements in scientific communities for accessing GPUs through grids or clouds. The GPGPU Virtual Team includes members from the Ireland, Bulgaria, Italy, Poland, Ukraine, United Kingdom, Spain, France, Netherlands NGIs, and from EGI.eu. These members developed the today opened two surveys - one for scientific communities, another one for resource administrators.
The purpose of the scientific communities survey is to gauge how the users currently use, or intend to use, GPUs or other accelerated devices in a distributed environment such as grids or clouds, or whether they shall use grid/cloud resources at all for this purpose. In particular, we would like to determine if there is a specific need from European research collaborations and their partners beyond Europe for a more tightly integrated GPGPU capability within the European Grid Infrastructure (www.egi.eu) or its Federated Cloud platform (go.egi.eu/cloud).
The purpose of the resource administrators survey is to gauge how the resource centre administrators (intend to) deploy and integrate GPGPUs or other accelerated devices into the Grid (or hybrid Grid/Cloud infrastructures), or whether they shall use grid/cloud resources at all for this purpose. In particular, we would like to determine if we can tightly integrate GPGPU capability within the grid/cloud environment in a consistent manner using established experience and best practice, or if there is a need from European research collaborations and their partners beyond Europe for a more tightly integrated GPGPU capability within the European Grid Infrastructure (www.egi.eu) or its Federated Cloud platform (go.egi.eu/cloud).
The deadline to fill the surveys is 13th of September, 2012. Information collected through the survey will be aggregated with information from other sites, and will be used by the relevant EGI teams to define a technical roadmap. Information will be also aggregated at a national level and provided to NGI representatives to allow the definition of NGI-specific action plans.
