Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The UAMS IT department does offer storage options for research data that might have identifiable data, such as patient information regulated by HIPAA, or student information regulated by FERPA.  Because of the potentially sensitive nature of such data, this storage is managed by UAMS IT to comply with various UAMS policies and procedures.  Identifiable data must be de-identified before it can be processed on Grace, since Grace does not comply with all UAMS policies for HIPAA and FERPA.  Naturally, data that does not fall under HIPAA or FERPA rules may also be stored in these systems. 

...

To reserve space on the Research NAS, or to mount a space that you have access rights to, please contact UAMS IT directly.  UAMS IT currently charges $1.50 per GB (subject to change) to reserve space on the Research NAS.  The reservation lasts 5 years.  After that. to continue holding that space a researcher would have to purchase another 5 year increment at the then going rate.  The charge is commensurate with the actual cost of purchasing the storage hardware needed to maintain the 3 replicas of the space, including the redundant disks for error coding that helps protect the data from hardware failures.

...

Box does provide means for sharing data, both inside and outside of UAMS, via cloud interfaces (e.g. through a browser).  Of course, a Box user should be careful that they do not share HIPAA or FERPA data inappropriately.  

Box provides the Box Sync Drive app for Mac and for Window apps that allows a user to sychronize access cloud based directory trees.  Box Drive optionally sychronizes locally stored directory trees with Box, giving a relatively simple and automatic method for backing up those directories, with the ability to use them offline.  In addition to Box SyncDrive, there are dozens of apps that can be used to organize and deal with data stored on Box on a variety of devices.

Please contact UAMS IT for information about setting up a Box space and about Box storage quotas.  

Storage Not Suitable for Identified Data

...

User do have the option to store data on their own systems, for example the local disk on their laptop or workstation, or on a local storage system (e.g. a departmental NAS).  In most cases, these storage systems are not being managed by IT in a fashion compliant with UAMS policies for HIPAA and FERPA, so care should be taken by users to not store identifiable or sensitive information inappropriately.  In general, sensitive data should use one of the UAMS-supplied storage systems as their primary home, and any copies that are synchronized or copied to local storage should be on protected, encrypted drives according to UAMS policies.  In addition, users are responsible for dealing with their own backup and disaster recovery needs. 

...

ROSS is visible inside the UAMS firewall, but is not currently directly visible outside of the UAMS firewall.  Essentially any system on the UAMS network has access to ROSS.  ROSS supports multiple protocols for storing and retrieving data, including S3, OpenStack Swift, NFS, and HDFS.  There are dozens of tools available that can move data to and from ROSS.  One of our favorites is Rclone rclone, which is universally available across multiple operating systems.  

The UAMS HPC team also maintains a data transfer node (DTN) that uses ecs-sync to move data between ROSS and Grace's cluster storage.  The ecs-sync program can move data faster and while using less compute overhead that Rclone can, but than rclone.  But some might find its ecs-sync's command line utility a bit more difficult to use.  Using the DTN is a good option for moving data to the Grace Cluster Storage prior to running a job that analyzes that data, and then again after the job completes to move the results and changed files back to ROSS.

...

OURRstore is an NSF funded project (circa $1 million) that is setting set up an Exabyte scale cold archive for research data.  We are eligible to use this storage resource for certain classes of research data.  It is a robotic LTO tape handler front ended by a disk-based cache system.  Access to data in OURRstore is through the cache system and a set of Data Transfer Nodes (DTNs).  LTO tape has an expected archive life of 15 years, with an error rate orders of magnitude lower that spinning hard drives.  Being conservative, the OURRstore team suggests planning on an 8 year lifetime.  The OURRstore system will always make at least 2 copies of data, one that stays in the robot, and a second that is sent back to UAMS as an offline, offsite backup of the data.  An optional third copy can also be made that is stored offline at a different location in Oklahoma.  This is the second LTO-based archive that the University of Oklahoma team has set up.  The first one (Petastore) has been in operation for almost a decade, and has proven extremely reliable and safe.

...

In addition to long, stable storage life, a major advantage to OURRstore is cost.  Since NSF funded the hardware and setup costs, and OU covers the administration and ongoing management costs, users of OURRstore only need to purchase media and pre-paid return shipping labels to put data in OURRstore.  At recently quoted prices, which continue to edge downward, that works out to under $20 per uncompressed TB for two replica copies, or $30 per uncompressed TB for triple replica copies kept at different locations, with an expected life of at least 8 years, possibly as long as 15 yearslonger.  Depending on the data, compression of 2x or more is typical, cutting those costs in half.  This is a tiny fraction of what cold archive in the cloud, such Amazon deep Glacial, costs.  This is the least costly option for stable, long-term storage of data.

...