The job class table of the PC cluster system for SX-ACE is as follows:
mode | job class | Maximum elapsed time | Maximum number of available Core | Maximum size of available memory | Maximum number of available nodes |
---|---|---|---|---|---|
shared use mode | ACE | 120 hours * HPCI users can run 1 node job for 500 hours |
1024Core (4 Core x 256 nodes) |
15TB (60 GB x 256 nodes) |
256 nodes |
DBG | 20 minutes | 32Core (4 Core x 8 nodes) |
480GB (60 GB x 8 nodes) |
8 nodes | |
INT | 60 minutes | 4 core | 60 GB | 1 node | dedicated use | myACE | unlimited | 4 core x number of dedicated nodes | 60 GB x number of dedicated nodes | number of dedicated nodes |
How to choose a job class
We provide two modes on SX-ACE.
Please submit to job class "ACE","DBG","INT" in the shared-use mode. Please submit to job class "myACE" in the dedicated-use mode. "DBG" is the job class for the debug in the shared-use mode. "INT" is the job class for an interactive batch in the shared-use mode. The details for the "interactive batch" can be found here.
Job Execution Class
Job requests submitted to job class “ACE” and “myACE” are categorized further into several “Job Execution Classes” . (The reason why the job class table is divided is caused by this mechanism).
In principle, users are not aware of the “Job Execution Class”, but the job execution class is displayed when users use scheduler commands such as qstat and sstat.
Job execution class in shared-use mode
Please choose the job class by the number of nodes you want and by HPCI or HPCI.
number of nodes |
job execution class | job execution class(for HPCI user) |
---|---|---|
256 - 129 nodes | A256C | A256H |
128 - 65 nodes | A128C | A128H |
64 - 17 nodes | A64C | A64H |
16 - 2 nodes | A16C | A16H |
1 node | A1C | A1H |
Under shared-used mode, users have to send the request for 512 - 257 nodes.
Please contact: system {at} cmc.osaka-u.ac.jp
Job execution class in dedicated-use mode
The job execution class is prepared for each group under the dedicated-use mode.
The job execution class for those who perform normal computation is A + (GID).
For example, the job execution class for a user with GID 14000 is “A14000″.
Limitations of Resource Usage
Each queue has an upper limit in terms of elapsed time, the number of CPUs available, and the number of nodes available, etc.
Users are required to specify their requests within the limitations in their job script file.
* In the case of requesting more than the limitation ...
We can change the limitation to some extent. The following form should be used for inquiries about limitation use.