- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -
- Information contained in the HLRS Wiki is not legally binding and HLRS is not responsible for any damages that might result from its use -
GPFS: Difference between revisions
No edit summary |
No edit summary |
||
| Line 1: | Line 1: | ||
'''IBM's General Parallel File System''' (GPFS) provides file system services to parallel and serial applications. It's a very fast and large file system which will be shared on following HWW hosts: | '''IBM's General Parallel File System''' (GPFS) provides file system services to parallel and serial applications. It's a very fast and large file system which will be shared on following HWW hosts: | ||
| Line 8: | Line 7: | ||
The file system uses the fast infiniband network infrastructure. The peak performance is about 800 MB/sec. A single infiniband node has a performance of about 200 MB/sec. The capacity of this gpfs is 15 TB. | The file system uses the fast infiniband network infrastructure. The peak performance is about 800 MB/sec. A single infiniband node has a performance of about 200 MB/sec. The capacity of this gpfs is 15 TB. | ||
A user has access from the listed nodes above only by using the workspace mechanism. The allocated workspaces will be shared by the nodes listed above. So the user has a fast access to the same workspaces from different PC clusters. <BR> | A user has access from the listed nodes above only by using the [[Workspace_mechanism|workspace mechanism]]. The allocated workspaces will be shared by the nodes listed above. So the user has a fast access to the same workspaces from different PC clusters. <BR> | ||
At the moment only some user groups have the permission to use the GPFS file system. To get a permission please ask the system administrators. | At the moment only some user groups have the permission to use the GPFS file system. To get a permission please ask the system administrators. | ||
Revision as of 14:55, 9 November 2007
IBM's General Parallel File System (GPFS) provides file system services to parallel and serial applications. It's a very fast and large file system which will be shared on following HWW hosts:
- cacau frontend and all cacau compute nodes
- phoenix frontend and all strider compute nodes
- volvox frontend and all volvox compute nodes
- ia64 frontend and all ia64 compute nodes
The file system uses the fast infiniband network infrastructure. The peak performance is about 800 MB/sec. A single infiniband node has a performance of about 200 MB/sec. The capacity of this gpfs is 15 TB.
A user has access from the listed nodes above only by using the workspace mechanism. The allocated workspaces will be shared by the nodes listed above. So the user has a fast access to the same workspaces from different PC clusters.
At the moment only some user groups have the permission to use the GPFS file system. To get a permission please ask the system administrators.