Weiterentwicklung der Linux Cluster Umgebung bei EDAG
The Engineering & Design A.G. (EDAG) Linux Cluster System was first started in Jan 2002. It consisted of 16 CPU’s running on the industry standard K7 Tyan hardware. The calculation turnaround times caused by the new Linux machines improved job turn around times so dramatically that the Linux cluster approach was seen as the future of High Performance Computing (HPC) for EDAG’s CAE hardware culture. It was just a matter of 12 months before EDAG had another 13 clusters each consisting of 16 CPU’s. These systems were located in groups at separate locations near the FEM engineering teams. The original management of these systems was a rudimentary scripting effort which through the LSF (queuing software) placed jobs on the clusters. The hardware management was facilitated through open source software. The road map of Cluster centralization was a pioneer effort which sought to bring multiple Linux cluster installations into one entity and govern it with one Job Management system, and to be nationally and internationally accessible for job submission and accounting. This was by no means a trivial effort which could be completed through over the counter software, but was a project that required the usage of self written ideas and concepts which were brought to bear on the maturing cluster environment.
https://www.dynamore.de/de/download/papers/forum04/weiterentwicklung-der-linux-cluster-umgebung-bei/view
https://www.dynamore.de/@@site-logo/DYNAmore_Logo_Ansys.svg
Weiterentwicklung der Linux Cluster Umgebung bei EDAG
The Engineering & Design A.G. (EDAG) Linux Cluster System was first started in Jan 2002. It consisted of 16 CPU’s running on the industry standard K7 Tyan hardware. The calculation turnaround times caused by the new Linux machines improved job turn around times so dramatically that the Linux cluster approach was seen as the future of High Performance Computing (HPC) for EDAG’s CAE hardware culture. It was just a matter of 12 months before EDAG had another 13 clusters each consisting of 16 CPU’s. These systems were located in groups at separate locations near the FEM engineering teams. The original management of these systems was a rudimentary scripting effort which through the LSF (queuing software) placed jobs on the clusters. The hardware management was facilitated through open source software. The road map of Cluster centralization was a pioneer effort which sought to bring multiple Linux cluster installations into one entity and govern it with one Job Management system, and to be nationally and internationally accessible for job submission and accounting. This was by no means a trivial effort which could be completed through over the counter software, but was a project that required the usage of self written ideas and concepts which were brought to bear on the maturing cluster environment.
E-II-1.pdf
— 601.5 KB