Storage

Several different locations are available for file storage for each cluster. Each filesystem has unique performance and backup characteristics and is optimized for certain kinds of use.

Locations

Location

Filesystem Type

Quota

Backup

Use Case

/home

Standard NFS

8 GiB

Yes

Config files, small scripts

/work/orion

Lustre

1.3 PiB (center-wide)

No

MSU user and project files, job I/O

/work/noaa

Lustre

Per project

No

NOAA project files, job I/O

/work2/noaa

Lustre

Per project

No

NOAA project files, job I/O

/scratch/orion

Lustre

None

Purged

Temporary files

NOAA projects are provided space under either /work or /work2 as determined by RDHPCS. To request quota increases, the project PI or manager should contact RDHPCS at rdhpcs.orion.help@noaa.gov.

MSU users may request space on /work. To request space, email help@hpc.msstate.edu with the cluster name, the user or project name, a justification, and the storage allocation size you need. MSU users are provided space under /scratch/orion/users automatically, however note that this space is subject to regular purges. Shared project scratch directories can be created upon request and with justification.

MSU user and project files follow the structure outline in this table:

Directory Type

Path

MSU Project

/work/orion/projects/<project>

MSU User

/work/orion/users/$USER

MSU Project Scratch

/scratch/orion/projects/<project>

MSU User Scratch

/scratch/orion/users/$USER

Location

Filesystem Type

Quota

Backup

Use Case

/home

Standard NFS

8 GiB

Yes

Config files, small scripts

/work/hercules

Lustre

2 PiB (center-wide)

No

MSU user and project files, job I/O

/work/noaa

Lustre

Per project

No

NOAA project files, job I/O

/work2/noaa

Lustre

Per project

No

NOAA project files, job I/O

/scratch/hercules

Lustre

None

Purged

Temporary files

NOAA projects are provided space under either /work or /work2 as determined by RDHPCS. To request quota increases, the project PI or manager should contact RDHPCS at rdhpcs.hercules.help@noaa.gov.

MSU users may request space on /work. To request space, email help@hpc.msstate.edu with the cluster name, the user or project name, a justification, and the storage allocation size you need. MSU users are provided space under /scratch/hercules/users automatically, however note that this space is subject to regular purges. Shared project scratch directories can be created upon request and with justification.

MSU user and project files follow the structure outline in this table:

Directory Type

Path

MSU Project

/work/hercules/projects/<project>

MSU User

/work/hercules/users/$USER

MSU Project Scratch

/scratch/hercules/projects/<project>

MSU User Scratch

/scratch/hercules/users/$USER

Location

Filesystem Type

Quota

Backup

Use Case

/home

GPFS

8 GiB

No

Config files, small scripts

/work/ptolemy

GPFS

Per Project or User

No

User and project files, job I/O

/scratch/ptolemy

GPFS

None

Purged

Temporary files

/reference/ptolemy

GPFS

None

No

Reference datasets for classes and research groups

Users may request space on /work. To request space, email help@hpc.msstate.edu with the cluster name, the user or project name, a justification, and the storage allocation size you need. Users are provided space under /scratch/ptolemy/users automatically, however note that this space is subject to regular purges. Shared project scratch directories and reference directories can be created upon request and with justification.

User, project, and reference files follow the structure outline in this table:

Directory Type

Path

Project

/work/ptolemy/projects/<project>

User

/work/ptolemy/users/$USER

Project Scratch

/scratch/ptolemy/projects/<project>

User Scratch

/scratch/ptolemy/users/$USER

Class Reference

`/reference/ptolemy/class/

Examples

`/reference/ptolemy/examples/

Other Reference

`/reference/ptolemy/

Location

Filesystem Type

Quota

Backup

Use Case

/home

VastFS NFS

30 GB

No

Config files, small scripts

/project

VastFS NFS

Per project

No

Project files, job I/O

/90daydata

VastFS NFS

None

Purged

Temporary files

/reference

VastFS NFS

None

No

Reference datasets

Project are automatically populated under the /project and /90daydata directories. To request a quota increase for an existing SCINet project, see the SCINet Request Resources page.

The /reference directory is a nightly sync from ceres.scinet.usda.gov:/reference/data.

Location

Filesystem Type

Quota

Backup

Use Case

/home

VastFS NFS

12 GB

Yes

Config files, small scripts

/work/morrill

VastFS NFS

Per project

No

User and project files, job I/O

/scratch/morrill

VastFS NFS

Node

Purged

Temporary files

Users may request space on /work. To request space, email help@hpc.msstate.edu with the cluster name, the user or project name, a justification, and the storage allocation size you need. Users are provided space under /scratch/morrill/users automatically, however note that this space is subject to regular purges. Shared project scratch directories and dataset directories can be created upon request and with justification.

User, project, and dataset files follow the structure outline in this table:

Directory Type

Path

Project

/work/morrill/projects/<project>

User

/work/morrill/users/$USER

Datasets

/work/morrill/datasets/center/<dataset>

Project Scratch

/scratch/morrill/projects/<project>

User Scratch

/scratch/morrill/users/$USER

Important

DO NOT run jobs in /home due to the small quotas and often poor performance for parallel I/O.

Important

Long term storage of data should not be on cluster storage as most directories are not backed up and some directories are subject to regular purges. Data should be moved off of the cluster when not in use.

Warning

DO NOT touch files to prevent purging of files. Attempting to circumvent purging is a violation of ARC policies.

For convenience, three environment variables are set on each cluster (except for Atlas):

Variable

Location

$HOME

/home/$USER

$WORK

/work/$CLUSTER

$SCRATCH

/scratch/$CLUSTER

Quota Management

Important

Avoid filesystem traversal commands (find, du -s *, etc.) on large directories. These commands can degrade performance for all users by consuming shared I/O resources.

Some storage locations have user-available mechanisms to check quotas.

Location

Command

/home

quota -s

/work/noaa

reportFSUsage

/work2/noaa

reportFSUsage -f /work2/noaa

For all locations not listed, there is not a user-available mechanism to determine quotas at this time. For all locations, if you need to identify files contributing to your quota or locate large files:

The support team can run these operations efficiently without impacting cluster performance.

Location

Command

/home

quota -s

/work/noaa

reportFSUsage

/work2/noaa

reportFSUsage -f /work2/noaa

For all locations not listed, there is not a user-available mechanism to determine quotas at this time. For all locations, if you need to identify files contributing to your quota or locate large files:

The support team can run these operations efficiently without impacting cluster performance.

Ptolemy does not have a user-available mechanism at this time. If you need to identify files contributing to your quota or locate large files, contact the helpdesk at help@hpc.msstate.edu. The support team can run these operations efficiently without impacting cluster performance.

Location

Command

/home + /project

my_quota

For all locations not listed, there is not a user-available mechanism to determine quotas at this time. For all locations, if you need to identify files contributing to your quota or locate large files, contact VSRC at help-usda@hpc.msstate.edu. The support team can run these operations efficiently without impacting cluster performance.

Morrill does not have a user-available mechanism at this time. If you need to identify files contributing to your quota or locate large files, contact the helpdesk at help@hpc.msstate.edu. The support team can run these operations efficiently without impacting cluster performance.

Further Reading

Topic

Description

File Transfer Guide

Moving data to and from the cluster

Lustre Filesystem

High-performance storage details