There are a small number of high-memory nodes, each equipped with 128 GB of RAM. Generally we do not recommend using these nodes unless your job requires more than 64GB of RAM on a compute node.
To use these nodes, add the special job parameter highmem. Here are some examples of running high memory jobs:
- Requesting 1 cpu core and 128GB of RAM:
- Requesting 4 cpu cores and 128GB of RAM:
-l h_data=32G,highmem -pe shared 4
Note that (the number of cores)*(h_data value) must not exceed 128GB. For example, “-l h_data=32G -pe shared 4” (32G*4=128G) and “-l h_data=16G -pe shared 8” (16G*8=128G) are ok. But “-l h_data=64G -pe shared 4” (64G*4 > 128G) will not run.
Be considerate to others: these few high-memory compute nodes are shared with other users. Please use them only when you have a job which requires more than 64G memory. When requesting highmem your job may experience a longer wait (queue) time before it starts.