We have seen such behavior on Solaris 8 64-bits

Suggest you add following in the /etc/rc3.d/S???   ( where you start your
tsm server )

LD_LIBRARY_PATH="/usr/lib/lwp:"$LD_LIBRARY_PATH;export LD_LIBRARY_PATH
LD_LIBRARY_PATH_64="/usr/lib/lwp/64";export LD_LIBRARY_PATH_64

and restart your tsm server.

We also noticed that your TSM server will perform better if

        you do not use unix level buffering for your TSM DB and TSM log (
forceio in /etc/vfstab)
        ( It seems unbelievable to us too but  internal benchmarks proved it
!!! )
        Your TSM db and log are  FileSystems ( not raw partitions !!! )


Have a look at follow FAQ on sunsolve

2957   Why do we have two set of thread libraries on Solaris 8?



-----Original Message-----
From: Broderick, Sean [mailto:Sean.Broderick AT IRL.XEROX DOT COM]
Sent: Wednesday, January 29, 2003 11:20 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: VERY HIGH %SYS CPU


Paul and others,

Some additional details as requested are:

(i) The stgpool is approx 440GB and spread across 31* 13.5GB LUNs and 3*
72GB internal fibre attached disks. Logs and database are on totally
separate LUNs. All LUNs originating from HDS9980V and average read and write
times are 10-11ms and 2.5 ms respectively.

(ii) No paging out, 8GB memory (~6GB being used by file cache and 1.2GB
allocated to application).

(iii) Network interface is gigabit with no errors reported and throughput
tested using nttcp (results up to 690Mbps achieved). Also flow control
enabled so no rx_overflow when multiple clients transmitting data.

T'is a mystery...

Thanks
Sean


-----Original Message-----
From: Paul Ripke [mailto:stixpjr AT BIGPOND.NET DOT AU]
Sent: 28 January 2003 23:01
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: VERY HIGH %SYS CPU


On Wednesday, Jan 29, 2003, at 00:11 Australia/Sydney, Broderick, Sean
wrote:

> Hi,
>
> The CPU usage (%sys) is extremely high, like 80 - 90%, on the TSM
> server
> (Sun V880 running TSM v5.1.5) during the backup window. Particularly
> when
> TDP R3 clients are attempting their SAP backups via backint / brbackup
> and
> as such the throughput is extremely poor (<9MB/s direct to disk cache
> via
> gigabit network).

How are your disk stgpool volumes configured? Is the system paging? How
much
RAM do you have? Any errors on your network interfaces? How fast does
an ftp
run?

Since the vast majority of the work done by dsmserv during backups is
I/O,
a high sys% CPU is to be expected. OTOH, our Sun TSM servers can hit 9
MB/s
on 100 Mb ethernet, with much older hardware.

Cheers,
--
Paul Ripke
Unix/OpenVMS/DBA
101 reasons why you can't find your Sysadmin:
68: It's 9AM. He/She is not working that late.
-- Koos van den Hout

________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service. For more information on a proactive anti-virus service working
around the clock, around the globe, visit http://www.messagelabs.com
________________________________________________________________________

________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service. For more information on a proactive anti-virus service working
around the clock, around the globe, visit http://www.messagelabs.com
________________________________________________________________________