In our quite simple environment, we are (still) doing direct backup to LTO, running Networker 8.2 on Linux (SLES11).
Does anybody know how we could properly the throughput per tape device over time, to see if the data rates match?
We tried sar, but this utility does not seem to address tape devices.
We found nsrio, but this unofficial utility is quite old.
nsrwatch has the relevant data, but cannot be redirected to a log file.
The statistics collected by Networker appear to be too inaccurate (only offering avarages over savesets or savegroups).
Should we try to measure what goes over the FC ports?
Thanks for offering a tip.
I think your best bet is to monitor transfer rates on switches for port on which tape is connected. If you use IO blade on library it gets complicated as you won't see individual values, but it might be worth checking if (in such case) library itself provides such report (some of them do).
There are a bunch of issues which could lead to shoe-shining. If you want to avoid most of them you need to run a backup from a locally (storage node) attached filesystem which you fill with a bunch of files (random pattern) with the defined file size. Then test the backup speed.
Next you repeat the test using the same amount of data but with file sizes 10 times smaller ... and so on.
The general question to answer is - how will these test help you? ... because you cannot replace your source data.
In general, you do not have a choice - sooner or later you must implement B2D and clone to tape.