Inter-Cluster File System "Common" is Now Live

HCC users have four file storage options. The best one to use will depend on your application and the desired performance of the file system. A larger version of this image is available at https://hcc-docs.unl.edu/display/HCCDOC/Handling+Data
HCC users have four file storage options. The best one to use will depend on your application and the desired performance of the file system. A larger version of this image is available at https://hcc-docs.unl.edu/display/HCCDOC/Handling+Data

Traditionally, users who wanted to run workflows on multiple HCC clusters had to manually transfer the data between resources. While Globus and the high-speed transfer nodes have done much to alleviate the headache associated with transferring files between clusters, users were still often forced to choose between using a cluster which offered the hardware configuration they needed and accessibility to files that could not be easily moved. This issue will forever be a thing of the past due to the new $COMMON file system.

In addition to their $HOME (/home/group/user/) and $WORK (/work/group/user/) directories, all HCC users now also have $COMMON (/common/group/user/) directories. This allows users to access files stored on $COMMON regardless of what cluster they are working on. The $COMMON file system is available on all of HCC’s traditional clusters; Crane, Tusker and Sandhills.

While $COMMON functions most like $WORK, there are a few fundamental differences. Quotas on $COMMON are individually based like $HOME but with larger quotas per user. However, unlike both $HOME and $WORK, additional storage can be purchased on $COMMON if necessary. Like $WORK, $COMMON is not backed up. But, unlike $WORK, $COMMON will not be subject to purging. To prevent file loss, users are strongly encouraged to backup files stored on $COMMON on a regular basis.

Like all file systems, $COMMON has its limitations. Please consider these best practices when using your new directories:

- Do not use $COMMON for high I/O workflows. Active job output should still be directed to $WORK directories for best job performance.
- Limit the number of files in your $COMMON directory. Storing excessively large numbers of files will cause slow access for all users.
- Do not expect your compiled program binaries to work everywhere! $COMMON is available on machines with different CPU architecture, different network connections, etc.

For information on how to access $COMMON and details on its use, please visit Using the /common file system. Any questions, comments and concerns are welcome and should be directed to hcc-support@unl.edu