Pacific Teck sponsored a PC Cluster Consortium (Tokyo) event and was invited to speak about job schedulers. We presented Univa Grid Engine’s capability to intelligently divide compute node resources, including GPUs, so that mulitple job can run simutaneously on a single node. We also reviewed refererence cases with Univa Grid Engine that Pacific Teck has been supporting. Some of the Univa Grid Engine eferneces that Pacific Teck has been supporting are the largest supercomputers in the world.
We also presented the integration between Univa Grid Engine and BeeOND scratch burst bufferlike file system. We also discuss the issues with integrating job schedulers with container systems such as Singualrity and Docker. We had general consenus in the room that supporting Singularity containers was important for the industry.