RIKEN AIP

Located in Japan, RIKEN AIP is a project at RIKEN dedicated to machine learning and artificial intelligence. RIKEN AIP utilizes the most advanced machine learning centric hardware from NVIDIA including 52 DGX-1 servers, provided by Fujitsu. Pacific Teck provided Univa Grid Engine to Fujitsu and RIKEN AIP in order to take full advantage of all the resources in these extremely dense servers. Univa Grid Engine makes it possible to allocate resources such as cores, GPUs, memory, and interconnect by job which is more efficient than other solutions. For example, if one job needed just 1 GPU but reserved a whole DGX-1 node, then 7 GPUs would go to waste. Univa Grid Engine allows more research to be done at RIKEN AIP by ensuring all resources are fully accessible.

For more information on the specific configuration at RIKEN AIP,
please visit our partner Fujitsu here:

http://www.fujitsu.com/global/solutions/business-technology/tc/fields/case-studies-e/riken/

 

System Overview

Total theoretical computing performance
  • 54 PFLOPS (FP16)
Calculation server
  • NVIDIA DGX-1 54 units
  • FUJITSU Server PRIMERGY CX2550 M4 64 units
  • FUJITSU Server PRIMERGY RX2530 M2 32 units
Storage system
  • FUJITSU Server PRIMERGY RX2540 M2 6 units
  • FUJITSU Storage ETERNUS DX200 S3 8 units
  • FUJITSU Storage ETERNUS DX100 S3 1 unit
Software
  • FUJITSU Software FEFS
  • Univa Grid Engine

 

 
 
Univa Grid Engine
Learn More about Univa Grid Engine