High Performance Workspace for Computational Clusters
High Performance Workspace is available on all HPC systems. Each user is allocated 500GB space by default in /hpctmp/[username]. Users can access the workspace from any login nodes. The High Performance Workspace can help to improve the overall data processing performance of your jobs, therefore, you are always encourage submit your jobs from these two workspace.
The workspace is designed to be used like a normal Linux directory. Hence, all Linux file management commands (like ls, cd, mv, cp, rm) and C and Fortran file I/O functions calls are applicable.
You can create your sub folder in /hpctmp/[username]/. The example on how you can access and use it by changing to this directory is as below:
| Linux Command | Description |
|---|---|
| # cd /hpctmp/a0123456 | change directory to your folder in /hpctmp |
| # mkdir myfolder1 | Create a sub-folder “myfolder1” |
| # cd /hpctmp/a0123456/myfolder1; pwd | change to your sub-folder in /hpctmp/, print full path |
You can copy files between your HPC home directory and workspace as below:
| Linux Command | Description |
|---|---|
| # cp ~/file1 /hpctmp/a0123456/ | Copy file1 from HPC home directory to /hpctmp/a0123456. |
| # cp /hpctmp/a0123456/file2 ~/file2 | Copy file2 from /hpctmp/a0123456 to HPC home directory. |
You can also list out simple instructions on using the /hpctmp work space via command “hpc hpctmp” in any HPC login node. You can check out the disk space usage versus disk quota using command “hpc s”.
Please note that these workspace is designed as temporary workspace, it is not recommended to use them for long term mass data archiving purpose. Files that are not in use should be copied back to your own PCs or local storage media for archiving, and then to be deleted from the workspace. Idle files (no read/access/write activity) longer than 60 days will be purged from the workspace.
To transfer files/folders between the local disk on your Windows computer and workspace /hpctmp, please follow this instruction (PDF) using the Secure Shell file transfer utility or other equivalent tools. For Max/Linux users, you can use the equivalent secure file transfer utilities or use scp command from the terminal.
A secondary high performance scratch space is provided in Atlas9 and Volta GPU cluster at /scratch2/[username]/, which is specially cater for AI and machine learning simulations to be performed in the Volta GPU servers. You can access and use the dedicated and faster /scratch2/[username]/ workspace via Atlas9 login node. There is neither backup nor snapshot for /scratch2 workspace, you are suggested to make necessary backup of data and results at your local storage media. Idle files (no read/access/write activity) longer than 60 days will be purged from the workspace.
To achieve optimum performance, backup (including snapshots) of data and files is not enabled in the workspaces. You are recommended to backup your data/files as-and-when it is necessary to avoid data loss.
To maintain the performance and maximize the usage, automatic purging of files older than 60 days is applied to both workspaces.
Each user is allocated 500 GB disk space in both /hpctmp and /hpctmp2, therefore, the total disk quota for each user is 1TB. Users are encourage to do regular house-keeping to prevent the disk quota being exceeded and your running jobs are affected. You can check the usage and disk quota via the following customized command:
# hpc space
--------------------------------------------------------------
Disk space for HPC home directory (quota -v)
--------------------------------------------------------------
HPC Home Dir Usage Quota Limit(KB) Status
/home/svu/ccev711 1221623, 3984384, 4194304, 30.66%
==============================================================
--------------------------------------------------------------
Disk space for HPC workspace at /hpctmp2
( lfs quota -u username /hpctmp2 )
--------------------------------------------------------------
HPC WorkSpace Usage Quota(KB) Limit(KB) Status
/hpctmp2 0, 512000000, 563200000, 0.00%
==============================================================
-----------------------------------------------------------------
Disk space for HPC WorkSpace
-----------------------------------------------------------------
HPC WorkSpace Usage, Limit, Status
/hpctmp/ccev711 8.86G, 500.00G, 1.77%
=================================================================
If any user has a justifiable reason for more disk quota for the High Performance Workspace, you can request for more disk space in one of them. As /hpctmp is currently less used and easier to expand, users are encourage to put up the request for /hpctmp.
To submit your request, please follow the steps below:
- Login to the Service Portal
- Choose IT Service Catalogue
- Scroll down and choose HPC Service Request
- Scroll down and choose HPC Workspace. Click on Quota Increase Request. Choose either /hpctmp or /hpctmp2, fill in the online request form and Submit.