Skip to Main Content

The University of Tennessee

Newton header banner

Frequently Used Tools:

Home » Documentation » Systems » Monster

Monster - 1TB shared-memory node

The "Monster" system is a single compute node with 1TB (1024 GB) of RAM. It can run up to 48 job threads all accessing the same 1TB shared memory space. This system is accessible by all Newton users via the short_monster and medium_monster queues. To request use of this system, a job must request the queue, the amount of RAM required, and the special job option "-l monster". Here is an example job file:

#$ -q short_monster
#$ -l mem=512G,monster

To request use of multiple threads or processes, the job can use the "cores_per_node" variable with values of 1, 4, 8, 12, 16, 24, and 48:

#$ -q short_monster
#$ -l mem=512G,monster,cores_per_node=16

You can also use "-pe threads N" to request N threads on the system, but you will need to modify your RAM request (-l mem=...) by dividing by the number of threads. Therefore, an equivalient request to the previous example is as follows.

#$ -q short_monster
#$ -pe threads 16
#$ -l mem=32G,monster

This example requests use of a total of 512GB of RAM and 16 CPU threads for this job. Note that this system does not support the use of "-pe openmpi".

Per-user limits on the monster system are as follows:

Queue CPU limit RAM limit
short_monster cores_per_node=48 mem=1024G
medium_monster cores_per_node=24 mem=512G

Back to Newton HPC Systems.