HPC-Scratch

Usage instructions

The ZID service called HPC-SCRATCH provides a consolidated file space to HPC users. It is shared by and thus accessible from currently two HPC systems, LEO1 and LCC.

The category of an HPC-SCRATCH account specifies its size and depends on a user's demand. You may contact us to adapt the file system size to any other category or, in execptional cases, adapt your file system size to an even higher individual size.

In the following we provide some recommendations on how to access the common scratch directory from the two different HPC systems:

Your HPC scratch directory includes a sub directory for both HPC systems, LEO1 and LCC.
Usage example for user "testuser":
```
[testuser@login.leo1 ~]$ ls /hpc-scratch/testuser/
lcc       leo1
```
HPC systems might differ with respect to architecture and instruction set. Both clusters, the AMD based LEO1 and the Intel based LCC, rely on the x86_64 instruction set. However, if you are using highly optimized code with, e.g., SSE instructions you might want to use different scratch spaces for different executables. In this case you can use the environment variable HPC_SCRATCH pointing to the cluster specific directory within your common scratch directory.
Issuing the command "echo $HPC_SCRATCH" on LEO1 results in:
```
[testuser@login.leo1 ~]$ echo $HPC_SCRATCH
/hpc-scratch/testuser/leo1
```
Issuing the same command on LCC shows:
```
[testuser@lcc16-be1t3 ~]$ echo $HPC_SCRATCH
/hpc-scratch/testuser/lcc
```
Each user's scratch file system is managed by the automounter.

This means: If you have not used your scratch space recently and issue "ls /hpc-scratch" or the like, the listing will not contain your directory (or may even appear empty). Just "cd /hpc-scratch/$USER" or use any other method to access a file or directory in "/hpc-scratch/$USER". Your scratch directory will be automatically mounted and visible, so you do not need to modify any jobs that access data in /hpc-scratch/$USER.