I was told by someone at netapp, that because of the nature of the WALF 
filesystem, block sizes greater then the current max, did not work out. I do 
not believe that they have changed this in 7.x

I can say that we have seen jobs running over 30,000. Also a number well below 
that rate. 
It has a number of depends, like the filer hardware, The ontap software level. 
The size of the object that one is backing up. The larger the better, as there 
is some overhead to starting up a ndmp dump job. 
Only doing one backup job per netapp volume. So if one has 10 qtrees in an 
netapp volume, then one should only back up one at a time.  Doing three at a 
time, each job will run less then 1/3 speed. 

Of course the classic backup things, like the more small files instead of large 
bulkly files will slow things down. 

The netapp does the backup, but all the detailed info on the job has to go to 
netbackup to store all that info in it's catalog. 

len

-----Original Message-----
From: veritas-bu-admin AT mailman.eng.auburn DOT edu [mailto:veritas-bu-admin 
AT mailman.eng.auburn DOT edu] On Behalf Of william.d.brown AT gsk DOT com
Sent: Monday, October 17, 2005 12:43 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: Re: [Veritas-bu] NDMP performance and tuning

The filer will send quite a lot of traffic back to the server with the NDMP 
option installed.  You want good IP connectivity and spare CPU power on the 
'Netbackup for NDMP' server.  There may be a few options on the 
filer, but I don't recall any way to set the buffer size.    NDMP 
appliances tend to be hard-coded at 64K, as that works with all tape drives, 
but obviously cannot make LTO reach top speed.

In fact I found this in the man pages of a filer:

DATA_BLOCK_SIZE     4-64    63  Specifies the tape blocking factor in 
kilobytes.
 
These are NDMP protocol options, in principle you can put them in the file 
list and they are passed to the NetApp.   But wow, you can reduce the 
block size..but the default is 63 blocks.  That is Data OnTap 6.5, I guess it 
might have changed in 7.  Not good!

William D L Brown




"Paul Keating" <pkeating AT bank-banque-canada DOT ca> Sent by: 
veritas-bu-admin AT mailman.eng.auburn DOT edu
17-Oct-2005 13:46
 
To
"Veritas List" <veritas-bu AT mailman.eng.auburn DOT edu>
cc

Subject
[Veritas-bu] NDMP performance and tuning






I'm currently doing some NDMP testing.
 
Netbackup 5.0MP4 master server, on Sun V240/Solaris 8 NetApp gf940 filer, SAN 
attached to Hitachi storage.
Adic Scalar 24 library with two IBM LTO2 SCSI drives and onboard FC bridge.
 
all devices are attached via a Brocade SW24k.
 
"Virtual Private SAN" utility on the Library FC brdige was used configured to 
advertise the robotic controller and one drive to the master server, and the 
other drive to the filer.
 
Brocade has once zone for filer and drive, and second zone the master server, 
library and 2nd drive.
 
NDMP backups have been running at 14,000, 16,000, 18,000 kbps........not in the 
30,000+ neighbourhood I was expecting.
 
Seems buffers and such on the master server would have *zero* effect on NDMP 
data rates.....filer options for NDMP are basically just "on" and "off".
 
is there any IP "heartbeat" between filer and master that coul/would affect 
data throughput?
 
Paul
 
 


_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu 
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu