Dan Langille schrieb:
> On 6 Oct 2007 at 10:41, Ralf Gross wrote:
>
> > How can I get information about how often a tape drive started/stopped
> > writing to tape? Is there a way to monitor the throughput of the SCSI
> > interface to the drive?
>
> On FreeBSD, I'd tell you to look at iostat. Output looks something
> like this:
>
> $ iostat
> tty twed0 cd0 sa0
> cpu
> tin tout KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s us ni sy
> in id
> 0 30 31.02 34 1.03 0.00 0 0.00 62.99 1 0.08 3 0 2
> 0 95
>
> iostat -w 1 will output one of those lines every second.
With a custom build kernel and the stap tool I was now able to monitor
the write performace of one LTO-4 drive (blk_wrtn/s = MB/s).
without spooling (snapshot of a backup of 163 GB data, taken every
second):
Device: tps blk_read/s blk_wrtn/s blk_read blk_wrtn
st1 890.00 0.00 56070.00 0 56070
st1 1560.00 0.00 98280.00 0 98280
st1 1546.00 0.00 97398.00 0 97398
st1 1556.00 0.00 98028.00 0 98028
st1 1537.00 0.00 96831.00 0 96831
st1 1578.00 0.00 99414.00 0 99414
st1 1591.00 0.00 100233.00 0 100233
st1 1581.00 0.00 99603.00 0 99603
st1 1548.00 0.00 97524.00 0 97524
st1 798.00 0.00 50211.00 0 50211
st1 1220.00 0.00 76860.00 0 76860
st1 764.00 0.00 48132.00 0 48132
st1 1561.00 0.00 98343.00 0 98343
st1 1518.00 0.00 95634.00 0 95634
st1 1555.00 0.00 97965.00 0 97965
st1 1582.00 0.00 99666.00 0 99666
st1 1589.00 0.00 100107.00 0 100107
st1 1565.00 0.00 98595.00 0 98595
st1 1503.00 0.00 94689.00 0 94689
st1 1596.00 0.00 100548.00 0 100548
st1 461.00 0.00 28980.00 0 28980
st1 1243.00 0.00 78309.00 0 78309
st1 1225.00 0.00 77175.00 0 77175
st1 166.00 0.00 10458.00 0 10458
st1 1529.00 0.00 96327.00 0 96327
st1 1561.00 0.00 98343.00 0 98343
st1 1519.00 0.00 95697.00 0 95697
st1 1585.00 0.00 99855.00 0 99855
st1 1593.00 0.00 100359.00 0 100359
st1 1578.00 0.00 99414.00 0 99414
st1 1608.00 0.00 101304.00 0 101304
st1 1529.00 0.00 96327.00 0 96327
st1 365.00 0.00 22932.00 0 22932
st1 1403.00 0.00 88389.00 0 88389
st1 388.00 0.00 24444.00 0 24444
st1 129.00 0.00 8127.00 0 8127
st1 1415.00 0.00 89145.00 0 89145
st1 1543.00 0.00 97209.00 0 97209
st1 1506.00 0.00 94878.00 0 94878
st1 1596.00 0.00 100548.00 0 100548
st1 1590.00 0.00 100170.00 0 100170
[...]
with spooling:
st1 1682.00 0.00 105966.00 0 105966
st1 1684.00 0.00 106092.00 0 106092
st1 1681.00 0.00 105903.00 0 105903
st1 1461.00 0.00 91980.00 0 91980
st1 425.00 0.00 26775.00 0 26775
st1 1414.00 0.00 89082.00 0 89082
st1 578.00 0.00 36414.00 0 36414
st1 1673.00 0.00 105399.00 0 105399
st1 1684.00 0.00 106092.00 0 106092
st1 1684.00 0.00 106092.00 0 106092
st1 1684.00 0.00 106092.00 0 106092
st1 1685.00 0.00 106155.00 0 106155
st1 1684.00 0.00 106092.00 0 106092
st1 1685.00 0.00 106155.00 0 106155
st1 1305.00 0.00 82152.00 0 82152
st1 970.00 0.00 61110.00 0 61110
st1 837.00 0.00 52731.00 0 52731
st1 1058.00 0.00 66654.00 0 66654
st1 1684.00 0.00 106092.00 0 106092
st1 1684.00 0.00 106092.00 0 106092
st1 1684.00 0.00 106092.00 0 106092
st1 1685.00 0.00 106155.00 0 106155
st1 1685.00 0.00 106155.00 0 106155
st1 1684.00 0.00 106092.00 0 106092
st1 1683.00 0.00 106029.00 0 106029
st1 847.00 0.00 53298.00 0 53298
st1 1431.00 0.00 90153.00 0 90153
st1 360.00 0.00 22680.00 0 22680
st1 54.00 0.00 3402.00 0 3402
st1 1432.00 0.00 90216.00 0 90216
st1 1684.00 0.00 106092.00 0 106092
st1 1685.00 0.00 106155.00 0 106155
st1 1684.00 0.00 106092.00 0 106092
[...]
It seems that the job with spooling enabled can feed the drive at a
higher speed, but in the the end both jobs had an average write speed
of 75 MB/s (no spooling) and 77 MB/s (spooling) The compression ratio
of this backup was ~1.45:1. The spool device is a large RAID5 which is
able to do seq. reads at ~160 MB/s. BTW: with LTO-3 I was able to
write to tape with the same speed. I'm not sure what the limiting
factor is at the moment. Differnt blocksizes didn't change anything.
LTO-4 needs a minimum of 40 MB/s to stream the data continuously to
tape. I'm still not sure if the drive has to stop very often during
backup, because there are some dropouts below 40 MB/s (even with
spooling). On the other hand, the drive has a 128 MB buffer which
should be able to cache these short dropouts. I'd like to avoid
spooling because it nearly doubles the time needed for serveral TB.
Therefore I'm still looking for a way to get the number of start/stops
that a drive needs while writing to tape.
Does anyone know a tool that is able to get this information from the
drive - in case HP drives do store this information somewhere at all.
Ralf
-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
