Do I need OURRstore Tape Archive Training?

Owners of an OURRstore account require an OU Supercomputing Center for Education & Research (OSCER) account to manage their data.  So if you are a Bates principle investigator (PI) using OURRstore to archive your STEM research data, you will need an OSCER account and will be required to take this training.

At present (4/22/2020), training is upon request with a frequency of 2-3 times per month.  However, it is anticipated with the inception of OURRstore and the increased issuance of  OSCER accounts for PIs across 28 states, the format of the present training will change to accommodate more frequent training sessions.

When you request access to OURRstore using the OURRstore@Bates access request form [Link to form coming soon!] , a member of the Bates OURRstore support team will work with you to coordinate your training.

What are the minimum and maximum file sizes allowed by OURstore?

Minimum file size is ~1 GB.  Maximum recommended file size is 10% of the tape media capacity.  Practically speaking, OURRstore is looking for a size that can be read in a minute (~200 GB, at present).

How will I transfer my archival data files from Bates to OURRstore?

Data transfers will go over the internet, and will behave like a local-disk-to-remote-disk copy.

OURRstore encourages people to use Globus instead of SFTP/SCP, because Globus is faster and has more features. However, SFTP/SCP file transfer is supported.

How long will it take for my file to upload or download?

For OURRstore, uploading will typically be faster than downloading, from your perspective as a user.

Upload Time

When you upload files to OURRstore, you’ll actually be uploading to the front end landing pad hard disk. Upload speed will be the slowest of:

(a) the speed of the disk that the original copy of the file is being uploaded from (which could be anywhere from 0.1 Gigabits per second from your laptop to X Gbps from Etna or Leavitt HPCC);

(b) the speed of the network that the file is being uploaded over (X Gbps at Bates, probably ~40 Gbps on OURRstore);

(c) the speed of OURRstore’s front end landing pad hard disk (expected to be ~20 Gbps, in aggregate for all uploads and downloads happening at the same time, so your individual file upload will probably be slower, more like a few Gbps).

Once you’ve uploaded the file to OURRstore’s front end landing pad hard disk, special software will automatically notice the file and copy it from disk to tape, without you having to do anything, and without you having to wait for that to happen.

Download Time

Download speed will be the slowest of:

(d) time spent waiting for a tape drive to become available (typically this is zero or very little);

(e) time spent preparing to read the file from tape (typically this is half a minute to two minutes, usually one minute);

(f) time spent reading the file from tape to the front end landing pad hard disk (typically 2.4 Gbps).

In fact, to download a file from OURRstore to your computer at Bates, steps (d), (e) and (f) happen first, then (c)/(b)/(a).

Can I make my data accessible to others?

Yes, stable “links” to data can be hosted in an institutional repository for access by others.

Because external requests for OURRstore resident data files will initially be transferred from tape to a front end landing pad hard disk (where files get sent to over the internet for migration onto tape, or come to when being retrieved from tape for remote download), download times could be significantly longer than for files retrieved from disk.

However, for popular datasets OURRstore will do disk caching: a requested file retrieved from tape will stay on disk until its disk space is needed for other files, no matter how long that is.

This means if a file were getting retrieved over and over, then it is likely a significant number of those retrievals wouldn’t involve tape at all, because the file would still be on disk. These “disk only retrievals” would therefore be significantly faster.

 

Should I create 1 or 2 copies of my data?

OURRstore requires dual or triple copies. So at a minimum, you will duplicate your data.

Duplicate Data: 1 tape for OURstore and 1 tape to be sent to you when the tape is full.

• Advantage: Cheaper
• Disadvantage: If the OURRstore copy fails (for example, if its tape cartridge breaks), then you have to ship your copy to University of Oklahoma along with a prepaid shipping label for University of Oklahoma to return your tape. There is also a minor risk of losing those data, if something goes wrong with the shipping.

Triplicate Data: 1 tape for OURstore, 1 tape stored at a separate physical University of Oklahoma campus location, and 1 tape sent to you.

• Advantage: If the OURRstore copy fails, there are no shipping costs for creating a new OURRstore copy.
• Disadvantage: More expensive

NOTE: OURRstore does not allow single copy, because of a requirement from the National Science Foundation’s grant proposal reviewers.

 

What tapes should I buy? Where can I purchase them?

Tapes can be purchased from any vendor, but buy ONLY THE FOLLOWING MODEL:

IBM model 38L7302 (6TB raw native capacity; 9TB raw in “LTO-7 Type M” mode).

Here’s ONE EXAMPLE of a relevant reseller ($61.00 on 4/20/2020 with free shipping):

https://www.walmart.com/ip/IBM-LTO-Ultrium-7-Data-Cartridge-38L7302/123900541

Here’s another EXAMPLE, for a reseller that can handle large quantities of tape cartridges:

https://www.tapeandmedia.com/ibm-lto-7-tape-ultrium-tapes.asp

 

Where should my blank tapes be shipped?

When you order your tapes, please have them directly shipped to:

How does one obtain a prepaid return shipping label at Bates?

Assume a shipping weight of 0.75 lb per tape.

Details for generating a prepaid, self addressed return label are coming soon! (4/22/20)

NOTE: Some of the text on this webpage is from, or based on, documents provided by the University of Oklahoma (OU), especially by Henry Neeman, Director of the OU Supercomputing Center for Education & Research (OSCER).