Last update: 02/20/2009

Archiving files on NCAR's new HPSS Storage System

The following is for frost users who are interested in archiving files on the new HPSS (High Performance Storage System), which is scheduled to be put in production on March 2, 2009. In preparing this document, we assume that you wish to archive data you have generated on frost, or data you need for running models or programs on frost. Archived data is defined as long-term, meaning persisting for years.

Archival vs. other storage

Archived files are contrasted with ones you may place on a file server. It usually takes longer to read or write archived files, whereas you expect faster response from a server. Also, the lifetime of an archived file is usually expected to be longer than that of a file residing on a file server. Files written to HPSS have no expiration date; they will remain on the HPSS system until they are explicitly deleted by the owner.

Maximum file size

HPSS files can be up to 1 Terabyte in size, although in practicality, HPSS files are easier to work with if you keep the size in the 10's of Gigabytes range.

Requesting an HPSS account

Users who would like to use HPSS need to request an HPSS account. You will need a Kerberos password to use HPSS, which will be set up for you as part of the HPSS registration process. You will also have your HPSS home directory created for you. In addition, you will be added to the "hpssusers" mailman list, and will be sent other information via email.

To request an account: TeraGrid users should send email to:

help@teragrid.org NCAR users, should contact CISL customer support at:

https://cislcustomersupport.ucar.edu/evj/ExtraView

Once you have been set up with an HPSS account, you can begin archiving files right away. At the present time (March 2009), the only supported user interface to HPSS is HSI, which is described below.

HSI (Hierarchical Storage Interface) usage

HSI provides a Posix-like environment with support for all of the usual commands such as ls, rm, mv, etc. Transferring files to/from HPSS is done via the "put" and "get" subcommands. Put and get make use of all of the built-in parallelism and optimization of the HPSS transfer mechanism, and are very efficient. We have seen transfer rates in the range of 80 to 100 MBytes/sec during our early HSI testing.

Detailed documentation and man pages for using HSI can be found at http://www.mgleicher.us/index.html/hsi/.

Feel free to look at this to learn more about all of the features and capabilities of HSI. There is also an hsi man page on frost that you can look at (type "man hsi"). But just to get started, here is some basic information and a few examples:

To start hsi, all you need to do is execute the command /opt/hpss/bin/hsi. If you haven't done a "kinit" command to get a Kerberos ticket, hsi will prompt you for your password (which must be your Kerberos password). HSI will then do the kinit for you. If you quit hsi and start it up again the same day, you will not be asked for your password again. But if your Kerberos ticket has expired, you will be asked for your password again when you start hsi.

When hsi starts up, you will get a prompt that looks like the following (using login name "user1" for these examples"):

[HSI]/home/user1->

You are now in your HPSS home directory (which will always be /home/user, where user is your frost login name). An "ls -l" command will show you all of the HPSS files/directories in your home directory. For a quick look at all of the options supported by "ls", type "ls -help". Other useful commands are:

[HSI]/home/user1-> pwd

displays current HPSS directory.

[HSI]/home/user1-> mkdir DIR1

creates a new HPSS directory.

[HSI]/home/user1-> cd DIR1

changes directory to DIR1.

[HSI]/home/user1-> mv file.1 file.2

renames "file.1" to "file.2" in the current HPSS directory.

[HSI]/home/user1-> rm file.3

deletes HPSS file "file.3". (Permanently removes the file, and it can't be recovered).

Class of service

To create a new HPSS file, you first need to know about the option that controls the number of tape copies. All files are first written to the HPSS disk cache, and then within a few hours they are migrated to tape. HSI supports the "copies=" command to tell HPSS how many tape copies of your files you want to create. We currently only support 1 or 2 tape copies (i.e. copies=1 or copies=2). The default is 1 tape copy (if no "copies=" command is given). This should be sufficient for most files.

For very critical files that have no backup elsewhere and can't be easily re-created, you may want to specify "copies=2" before doing the "put" commands.

Example:

[HSI]/home/user1-> copies=2

makes two tape copies of files.

All "put" commands will now result in two tape copies being made. This remains in effect until you say "copies=1", or until you exit hsi.

Put subcommand

The simplest "put" command is just:

[HSI]/home/user1-> put file.01

writes local file "file.01" in the current working directory (on frost) to HPSS as "/home/user1/file.01".

For a different name on HPSS:

[HSI]/home/user1-> put file.01 : new.hpss.file

writes local file "file.01"to HPSS as "/home/user1/new.hpss.file".

You can of course use absolute path names for local or HPSS files. The local file always comes first (both put and get), before the colon.

Get subcommand

Examples:

[HSI]/home/user1-> get file.01

reads HPSS file "/home/user1/file.01" into the current working directory as "file.01".

[HSI]/home/user1-> get file.01 : new.hpss.file

reads the HPSS file "/home/user1/new.hpss.file" into "file.01".

To exit from the HSI environment:

[HSI]/home/user1-> quit

For more detailed information on all of the HSI commands, features, etc., see the online user guide or look at the hsi man page on frost.

Next page | Table of contents - Frost user guide

If you have questions about this document, please contact us by any of the means shown on our Customer Support page.

Address of this page: http://www.cisl.ucar.edu/docs/frost/hpss.jsp