Types of storage
We have three types of storage available to IBM SP users: DFS, GPFS, and HPSS.
- DFS (Distributed File System) includes each user's home directory. This is online disk intended for permanent, frequently used storage and is backed up on a daily basis.
- GPFS (General Parallel File System) is a large fast disk area intended for global access to temporary storage, for example, to hold output generated by your job. It is not currently backed up. Our intention is to maintain GPFS at no more than 50% full so there is always a large amount of scratch space available for executing jobs. Users should move GPFS files which are not being actively used to HPSS.
- HPSS is the archival storage system. It consists of a terabyte (TB) of disk cache and some 150 TB of robotic tape storage. HPSS may be accessed by ftp, parallel ftp, or HSI (hsi).
Projects should not duplicate executables or datasets in each user's area but should set privileges to share a master copy. Contact the consulting group for help in setting up a Unix group to facilitate sharing. Files not in immediate use, such as source code and output datasets, should be moved to HPSS.
Space allocations
- A research account is entitled to 500 MB (0.5 GB) on DFS. With sufficient justification, it is possible to have 1 GB. For projects with special requirements, we will consider extensions of this limit on a case-by-case basis.
- Data on GPFS should not be retained for more than one week and should be migrated to HPSS as soon as the files are not being actively used. Because of the working-directory character of GPFS, we have not yet set a space limit for GPFS usage. If we find that users are frequently leaving files on GPFS for more than one week, we may have to restrict GPFS usage to ensure that we have enough disk space available to run large jobs.
- Space on HPSS is for files which are not needed immediately. We have not yet set a space limit for HPSS usage. Even though HPSS is a very large storage system, space is not unlimited. Users must not store files unrelated to their NEESC projects on HPSS. They must also review their files periodically and remove unneeded files.
Consequences of abuse
Storage usage will be monitored periodically. If time permits, first offenders will be warned to clean up their space. Ignoring these warnings will result in loss of access privileges.
If a file system runs out of or gets low on space, user data on that file system will be subject to involuntary deletion without prior notice under the following guidelines.
- GPFS files which have not been accessed in the current week (seven-day rolling window) will be deleted.
- Files to be deleted will be selected randomly; there will be no attempt to guess which files are more valuable to the user.