[Veritas-bu] L180 Fibre LTO2 throughput survey
2003-11-04 11:22:20
I don't think anyone can tell you if you should see a certain throughput
for LTO2, as it depends so much on your disks and your data.
The kind of things to check....
Start a backup using the bpbkar32 -nocont option - you should find it
referenced in the manuals. A good manual is one that is all about tuning
NBU on Windows, but the principles are quite transportable. Using
-nocont will run the job but send no data to tape. You can see how fast
it goes in *getting* the data off disk. If that is way faster than you
get to tape, you may have some improvement to make at the tape end. If it
is not, then you may need to look at how fast your disks are going, or at
multiplexing to get the data arriving fast enough at the tape drive, or at
network buffer sizes etc.
You can look in the bpbkar and bptm logs for the wait & delay counters -
again that windows document explains those, and how to work out how much
of your backup duration is spent waiting for buffers. You can see if bptm
is waiting for full buffers or bpbkar for empty buffers. You can then
try increasing either the size of buffers or the count. Note that if you
are using Windows some drivers may require adjustment to allow you to have
buffers above 64K.
You can test using known data. On the HP web site are tools to create
standard data sets and do various other things. Look for their Ultrium
tape drive, and there is a link to a page of utilities. If you do a test
using data that is 1:1 compressible (i.e. not!) and say 1MB files you are
looking for 30MB/s (35MB/s for IBM LTO 2).
You can use the Veritas data analysis tools to see what your file sizes
actually are. If you have large numbers of small files, it will slow
down.
So many things....all that is certain is that it should be faster than LTO
1, if you can get the data to the drive fast enough.
|
|
|