Topics Map > Services > Research Computing and Support > CCAST

Data Management for CCAST Users

Key information and best practices for storing, transferring, and managing data on CCAST systems

    Introduction

    This document is intended to provide an overview of data management on CCAST systems, including data storage locations and quotas, data transfer tools, and best practices for managing data.

    Data storage

    There are three primary locations where users can store data on CCAST systems: home, scratch, and projects. Each location has a different purpose and a different limit (quota) on how much data can be stored there.

    The home directory is intended for personal storage and has a quota of 200 GB. Each user has their own home directory where they can store source code, scripts, and small data files. Home directories are backed up nightly, making them a safe location to store data long-term.

    The scratch directory is high-speed storage intended for temporary files related to active computations, and has a quota of 20 TB. Each user has their own scratch directory. Data stored in scratch is NOT backed up, and files in scratch older than 60 days are deleted automatically to maintain performance and availability.

    Project directories are intended for long-term storage of data, are backed up regularly, and have a default quota of 1TB. A project directory is shared among all users in the owner's project group, and is therefore an ideal place for sharing data within a lab, research group, or other research collaboration. Faculty project owners may request an increase in their project directory quota by completing the form here. All increases are at the discretion of the CCAST Executive Director.

    Data storage location quick reference

    The table below summarizes key information about each of the available storage locations mentioned above.

    Location

    Path Quota Ownership Backed up?

    Home

    Thunder: /mmfs1/thunder/home/<user>

    Prime: /mmfs1/home/<user>

    200 GB Private per user Yes

    Scratch

    Thunder: /mmfs1/thunder/scratch/<user>

    Prime: /mmfs1/scratch/<user>

    20 TB Private per user No

    Projects

    Thunder: /mmfs1/thunder/projects/<project_owner>

    Prime: /mmfs1/projects/<project_owner>

    1 TB Shared within project group Yes

    Checking data usage

    You can see your data storage quotas and usage at any time by using the lsquota command in the terminal, or by selecting "Disk Usage" from the "Utilities" menu in the CCAST OnDemand web portal (see screenshot below). 

    Location of the disk usage option on CCAST OnDemand

    When a storage quota is exceeded, the user will be unable to write to the directory until the amount of data stored is brought back under the quota (i.e., through deletion or migrating files to another location).

    Data transfer

    There are several tools available for uploading/downloading data to/from CCAST systems. Some of these are hosted by CCAST and others are third-party tools.

    Transfer tools

    CCAST OnDemand

    For data transfers less than 10 GB, files can be uploaded and downloaded using the CCAST OnDemand web portal. Upload and download buttons are available from any storage location in the "Files" menu (see screenshot below).

    File transfer in CCAST OnDemand

    Globus

    Globus is a data transfer service designed for high-performance, reliable, and secure data transfer. For CCAST users, Globus provides a convenient way to transfer data between CCAST storage locations, external storage systems, and other research institutions. Globus is especially well-suited for transferring large datasets. For detailed instructions on how to use Globus on CCAST, please see the Globus KB article.

    SCP

    SCP (Secure Copy Protocol) is a command-line tool for securely transferring files between a local host and a remote host or between two remote hosts. SCP uses the SSH protocol for data transfer, providing encryption and authentication mechanisms to ensure data security. SCP is widely used for transferring files across networks generally, and is supported by Windows and most Unix-like operating systems.

    For example, to copy a file from a local computer to CCAST Prime, the following command can be used:

    
    # A windows example
    > scp C:\path\to\file.txt <user>@prime.ccast.ndsu.edu:/path/to/destination
    # A unix-like example (Linux, MacOS)
    $ scp /path/to/file.txt <user>@prime.ccast.ndsu.edu:/path/to/destination
    

    The -r flag performs SCP recursively, allowing for entire directories to be copied. For instance, moving a directory from CCAST to a local computer:

    # Moving the directory to the current working directory
    $ scp -r <user>@prime.ccast.ndsu.edu:/path/to/directory/ /path/to/destination
    

    WinSCP

    WinSCP is a desktop application for Windows that allows users to interact with remote host systems using a more familiar drag-and-drop interface. After logging into CCAST via WinSCP, users will see a split-pane file explorer with their local computer on one side and CCAST on the other (see screenshot below).

    WinSCP between local computer and CCAST

    Rsync

    Rsync is a command-line tool for efficiently transferring and synchronizing files, and has the ability to resume interrupted transfers. Rsync is widely used and is supported by most Unix-like operating systems, and is especially useful for transferring large datasets or for incremental backups.

    For example, to copy a file from a local computer to CCAST Prime, the following command can be used:

    $ rsync -avz <user>@prime.ccast.ndsu.edu:/path/to/file.txt /path/to/destination
    

    Here, the -a flag preserves file permissions and ownership, the -v flag enables verbose output, and the -z flag compresses data during transfer.

    Partial transfers can be resumed by adding the --partial flag to the rsync command, and the -r flag performs operations recursively. For example:

    # Resume a partial transfer
    $ rsync -ravz --partial /path/to/directory <user>@prime.ccast.ndsu.edu:/path/to/destination
    

    Another useful set of flags is -bu, which will produce a backup and update files only if the source file is newer than the destination file:

    # Backup and update files
    $ rsync -bruvz <user>@prime.ccast.ndsu.edu:/path/to/directory /path/to/destination

    Data management

    Data management is a broad topic, so here, we focus specifically on best practices for managing data stored on CCAST systems.

    Where to store different kinds of data

    The different storage locations on CCAST are intended for different purposes.

    User home directories are private to each user and have a limited amount of storage space. As such, they are ideal for storing relatively small data that users want to remain private. Types of data that are ideal for storing in the home directory include scripts, code files, custom software (e.g. Python or R packages, executables downloaded or built from source), and small test data sets. Home directories are not ideal for large data or data that you want to be shared with other users on CCAST.

    User scratch directories are also private, but have the capacity to store much larger quantities of data. However, scratch directories are not backed up, and files older than 60 days which are not actively being used may be deleted to maintain the space of the system. For these reasons, scratch is ideal for storing large quantities of temporary or in-process data, data associated with computations, or databases for testing. Scratch is not ideal for data that you want to store long-term, or for data you want to share with other users on CCAST.

    Project directories can also store larger quantities of data than home directories (1TB and up) but are owned by faculty members and shared with all members of that faculty member's project group. Like home directories, project directories are backed up and can be used to store long-term data. Therefore, project directories are ideal for storing datasets, code, software, and other data that cannot fit in home, and that needs to be accessible by all members of the research group. Project directories should not be used to store data that users want to remain private, since other group members may be able to see data in the project directory.

    Tips for faculty project owners

    Faculty project group owners should have their students store research data in their project directory. There are several benefits to standardizing on this policy:

    • All data stored in a faculty member's project directory belongs to that faculty member, and will not be deleted or removed when students and staff depart from the lab. This provides for data access continuity.
    • Data stored in user's home directories is private, and faculty should not expect that they will be able to retrieve such data from one of their students' home directories upon departure. If faculty suspect important research data is being stored in students' home directories, they should remind them to move it to the project directory, as recommended above.
    • Data stored in project directories is accessible by all members of the project group. This makes project directories ideal for sharing data between different users, especially if they are collaborating on various projects in the same lab.

    Data management plans for grant proposals

    Many grant programs require a data management plan. CCAST is happy to provide letters of support and facilities descriptions for grant proposals, but cannot write the data management plan itself. If you plan include CCAST as part of your data management plan, and you anticipate that additional data storage and management resources will be needed beyond what you currently have access to, please talk with us before submitting the proposal. This will ensure that appropriate capacity is available in the event that the proposal is funded. In some cases, additional funds may need to be included in the proposal budget for the purchase of storage capacity, depending on the needs of the research.



    Keywords:
    CCAST, hpc, data management, data storage, data transfer mgmt 
    Doc ID:
    136595
    Owned by:
    Stephen S. in NDSU IT Knowledge Base
    Created:
    2024-04-04
    Updated:
    2024-06-10
    Sites:
    NDSU IT Knowledge Base