Well, to localize the problem I would follow these 2 steps :
1) disable directive set at client level. I often saw compression directive
that requires more client-CPU combined to anti-virus software to cause
performance drop-down. Check also for disabling verbose information
from savegrp.
2) create a FILE (or AdvFileDev) device and label it to a specific pool.
Force a full backup to that pool. It would eliminate tape device
part of the problem.
You could also use bigasm to generate dummy data from backup server
(and/or from client) to determine how far the tape drive can write wihout
reading from disk.
I would also have a try to fill /etc/hosts with clients addr and disabling
DNS name resolution from server side. NetWorker is an heavy dns query user
and if you dns got in troubles, then you can also obtain serious
performance degradations.
Cheers and HTH,
Thierry
> Just to recap,
>
> Network = 1G that means max 80GB/s towards the server.
> San to S-AIT(234?) what is the max speed there both comressed and
> uncompressed
> (can be tested with a local large file)
> What kind of clients also 1G net, what kind of data (large files, many
> small)
>
> All those factors influence the client speed and therefore the backup
> speed.
>
> In my environment Sol9 E450 2x400Mhz CPU 1 GB mem, I see the highest
> throughput on cloning as that does not use any network at all. LTO-2 on
> LSi320 scsi gives me 40M/sec on average while normal backups vary strongly
> according to the data on the client but (as my network is basically 100M
> full
> duplex) never more than 10M/sec.
>
> Hope this helps a bit,
>
> Maarten
>
> On Thursday 05 May 2005 17:40, Stan Horwitz wrote:
>> We're running our NetWorker server on a Sun v480 with 2GB RAM and two
>> CPU
>> boards in it. We are seeing performance (i.e., throughput) that is lower
>> than what I would expect, even with doing backups of servers on the same
>> subnet as our NSR server. The GB network port on the server is nowhere
>> near capacity and web have individual fibre lines between each of our
>> nine
>> Sony S-AIT drives and our backup server.
>>
>> I am wondering if anyone on this list has any suggestions on how I can
>> tune Solaris 9 to allow for greater throughput.
>>
>> --
>> Note: To sign off this list, send a "signoff networker" command via
>> email
>> to listserv AT listserv.temple DOT edu or visit the list's Web site at
>> http://listserv.temple.edu/archives/networker.html where you can
>> also view and post messages to the list. Questions regarding this list
>> should be sent to stan AT temple DOT edu
>> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
>
> --
> Maarten Boot,
> Compuware Europe B.V.
> Hoogoorddreef 5
> 1101 BA Amsterdam
> Tel: +31 20 312 6511
>
> --
> Note: To sign off this list, send a "signoff networker" command via email
> to listserv AT listserv.temple DOT edu or visit the list's Web site at
> http://listserv.temple.edu/archives/networker.html where you can
> also view and post messages to the list. Questions regarding this list
> should be sent to stan AT temple DOT edu
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
>
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listserv.temple DOT edu or visit the list's Web site at
http://listserv.temple.edu/archives/networker.html where you can
also view and post messages to the list. Questions regarding this list
should be sent to stan AT temple DOT edu
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
