How to recognize shoe shining on lto2 tapes

[Walter Kießl]

Hi,

I have a HP LTO2-FC-Tape attached to my hpux11.11 box (acutally it's a Legato Networker Backupserver with a STK L700 Tape Library connected via FC).

What I want is to see if the data transfer rate of our backup server is high enough to keep the tape(s) streaming.

Is there a way to gather some statistics (e. g. on the drives) which show me when the drive has to stop/start because of an insufficent amount of data provided?

Thanks for your help!
Walter

[Marino Meloni_1]

You can install L&TT, there is a functionality called device performance that may check the path from server to tape, and test called "device perfomance" that allow you to test the server itself. The tool will issue a result that let you know the max performance of the system

[Peter Brewster]

Hi Walter,

It is more subtle than that. Two cautions: First, LTO-2 drives are "adaptable" so even without shoe-shine you may be taking a performance hit. My observation is that there is almost a 2:1 forgiveness range. The cost of forgiveness is usually modestly reduced data capacity on a full cartridge.

Second, be very careful when doing tests. Many test situations, home-grown or otherwise, sometimes end up with either highly compressible data (as in, make the data by repeated copies of the same data set), or random or pseudo-random data that does not compress at all.

If a drive is shoe-shining you can probably hear it. The sound is similar to that at volume recognition time. Your library will have to be "not busy" and you may need a stethoscope or listening tube. When you hear a drive in shoe-shine - you'll know it. (From time to time you will hear reversal points also. They are normal.)

An experiment you might try. Turn compression on the drive off. Then do your usual back-up. From this observation you can go a long way towards calculating drive and/or server performance.

More interesting is what the write data stream looks like and what you can do to change it for the better in some respect, or make it worse.

There is performance data available from the drive. I do not know of any way to get it from HP tools however - and HP no longer distributes the information needed form the commands and decode the results, even if you have the capability.

Finally, some switches can tell you the gross data rates. Not much but it may be useful.

Peter Brewster
peter.brewster@comcast.net

[Walter Kießl]

Hi,

thanks for the input. I'm aware of the problems with using testdata or calculating data transferrates from switches or other performance statistics. Especially if you keep in mind that new drives have a lot of cache to buffer slow datastreams.

This is why I'm looking for some data which are derived from our 'real' backups. In my opinion statistics gathered directly from the drives would be the best indicator for tape start/stops in the backupstreams. And by the way I think it's much more convenient than listening to the drives all nights :-)

Again, thanks for your help.