Cloning performance low

[RoyG]

Hello,

Every day we are cloning the savesets of the last 24 hours with a script. For some save sets the performance is very low, writing at 200 Kb/s or less. Others are written at 3000 Kb/s. What can cause this performance problem for some save sets?

We are using a storagetek 9730 jukebox, with 2 DLT 7000 drives and Legato 6.0.2 on NT4 SP5.

Greetings Roy

 

[packageman]

I'm using the same ST hardware as you, but 6.1.1 on Solaris8. Previously, I was cloning about 42GB in about 10 hours, using the nsrclone command in a Perl script. I made some changes (see below), and now get about 10GB per hour! According to the man page, the options I stuck on mminfo to get the ssids(saveset IDs) make them come out in "written to the tape order" of the savesets I want to clone. (Let me know if you need a translation of this stuff.)

mminfo -o mo -r 'ssid,volume,fragflags' -q 'level=full,valid,savetime>$date,savetime<&quot;$date 23:59:00&quot;'|grep -v clone|grep -v ssid|awk {'print
$1'} >$sidlf

nsrclone -b <clone_pool_name> -S -f $sidlf

 

[spaixpert]

try to set the block_size to 0.

 

[fernandezd]

I usually clone 24 GIG from DLT to disk and this is fast.
This is done in less than 1 hour.

So if you have little to clone then it may be a good idea to get a bunch of disks add it as a device and clone to it.

Restore times for a 20 GIG DB is 18 Mins from file system.

FernandezD
At your service.
Unix systems Admin

 

[cptb]

Cloning speed is depending on how your savesets are multiplexed on Tape. No mather if you're doing savset or volume cloning. The nsrclone process reads the data savset by savset and if you have a lot of savesets multiplexed on one volume it might have to move around on the tape a lot to de-multiplex the saveset.