Access to storage resources is obtained through investments in UFIT Research Computing (made either as hardware investments or service investments). Investments are made in fixed units appropriate to the given resources. When a request to make an investment is made by submitting a hardware or services request form, we try to provision the resources within two business days.
Details regarding the implementation and use of UFIT Research Computing storage can be found on the Storage Use policy page.
NFS-based storage on HiPerGator systems is typically auto-mounted, meaning it is dynamically mounted only when users are actually accessing them. For example, if you have an invested folder named /orange/smith, you must specifically type in the full path of “/orange/smith” to be able to see the contents and access them. Directly browsing /orange will not show the smith sub-folder (unless someone else is using it coincidentally). Auto-mounted folders are common on HiPerGator systems, and include /orange, /blue, and even /home.
High-performance storage for research data
UFIT Research Computing offers two types of high-performance storage, each existing for different and distinct use cases. They are not interchangeable; investors should carefully consider which storage type their research project will require before making an investment. Both types may be needed to effectively support a project.
- The orange storage type is less expensive, offering lower relative performance than blue storage. It is based on disk drives with higher density, which increase capacity at the cost of speed. However, performance is still higher than the typical NAS appliance purchased from the consumer market, and is designed for long-term use in enterprise and research computing environments. This storage is ideal for keeping large or numerous files for long periods of time, with regular access for reading and writing.
- The blue storage type is more expensive, offering high performance for large groups of independent tasks running on HiPerGator to read and write simultaneously, including coordinated tasks running MPI programs. Allocation of scratch storage[link] (storage of data that only exists for the duration of a serial or parallel job) are made on this type of storage. Large files that are accessed for reading and writing in a random way by one or many tasks should also be stored on the blue storage type.
Storage for regulated data for research
Restricted Replicated Storage Units (RRSU) are provided for use in research projects requiring storage that is compliant with certain laws, regulations, or contractual requirements, such as HIPAA, FISMA, ITAR, EAR, or CUI. Data placed in RRSUs is encrypted and and can be accessed only within the ResShield and HiPerGator-RV systems, which require multiple factors of authentication for user access. Data replication occurs automatically, in order to prevent data loss from drive failure or other unexpected system events.
This storage type is available within both the [link]ResShield and HiPerGator-RV restricted data environments; view specific compliance information on the Regulatory Compliance[link] page.
Home directories
The system provides home directories for users on a small file system that is not intended for high performance work. Your home area is intended for source code, scripts, and other human-readable data. The amount of data in the home area should be kept to a minimum. A quota of 40GB is set for home area storage.
Programs writing to the home area run inefficiently and make the home area sluggish for all users. For both batch jobs and interactive jobs, all input/output operations (I/O) should take place within a group’s storage investment, as these file systems have been designed for high I/O activity.
Home directories do offer the advantage of having one week of daily snapshots available, which can be accessed by the user to recover older versions of files or accidentally deleted files.
Snapshots of user’s home directory files are stored on the file system. For more information concerning the use of snapshots, please see the RC Wiki page about Snapshots.
Storage quotas
Storage is allocated by quota assigned to groups, which is shared among the users in a group. When the quota limit is reached, no more data can be written. To prevent any crises, group members will receive email notification when their group’s storage exceeds 90% of their quota limit. The notification contains a list of all the members in the group, as well as the amount of storage used by each group member. Groups nearing their quota limit can handle this issue by deleting unneeded data, or by purchasing additional storage.
Temporary storage for burst data
Some investors may run jobs that create large amounts of temporary data, which at the end of the job is deleted. Such investors have the option to request a burst allowance for their storage investment. The size of this allowance is determined by UFIT Research Computing on a case-by-case basis, because it is dependent upon the software and workflow used by the research group.
A burst allowance gives the group the ability to exceed their normal storage investment quota for up to 30 days. Once the group exceeds their quota, they have 30 days in which to reduce their storage usage to an amount within their normal quota limit, at which point the burst allowance timer is reset. If the amount of storage used by a group has not been reduced below their quota limit after 30 days, the group will not be able to write data until fulfilling the requirement.
Data Protection
Investors can purchase data protection by selecting one of the following data backup services. These services are charged separately from UFIT Research Computing compute services. The most cost-effective way to use these services is for the research group to select a single directory for backup, and to place all data to be backed up within the chosen directory.
Data protection with tape
UFIT Research Computing has partnered with UFIT Infrastructure Communication Technologies (ICT) in order to provide their Tivoli Storage Manager (TSM) tape backup service for data stored on UFIT Research Computing systems. The cost for data storage on tape is $78/TB/year. As such this service is billed differently than UFIT Research Computing provided services, in that clients are charged for usage rather than purchasing a certain amount of capacity. Investors must set up an account with UFIT-ICT to pay for the monthly charges,
After a research group has chosen a directory for backup, tape backups will occur once per day in incremental fashion – for each file that has changed since the previous backup, a new copy is written to tape. Up to seven versions of the same file can exist, and individual versions kept for 60 days. As long as the file exists on disk, at least one version will be kept on tape. If the file is deleted, the last version on tape will be removed after 90 days. For more information about the in-depth backup options available using this system, please see our wiki page.
If you are interested in tape backup for your group, please notify us by opening a support ticket.
Data protection on home directories
Home directories offer the advantage of having one week of daily snapshots available, which can be accessed by the user to recover older versions of files or accidentally deleted files. More information about the home directory snapshots can be seen on our wiki page about snapshots.