How to reduce CPU load on TSM server during backup window

[bonobos]

Hi,

how can I do a "tuning" of TSM that the CPU isn't jumping to nearly 100% during the backup window of several servers? (backup of 20 servers starts at 2 am). Should I split the starttimes to 2 and 4 (each with 10 servers)? Would that help?

Or are there some parameters in the TSM configuration which could help me to better utilize the resources?

What exactly does the TSM client acceptor deamon, is it more usefull than just the normal TSM client scheduler?

Thanks for some help!

 

[chad_small]

You need to provide more information:

What type of server is it? Backplane matters since TSM is a heavy I/O product.
How many CPU's in the TSM server (How many cores? What Speed are they too?)
What OS is the TSM server?
How much RAM?
What is the LAN connection between the server and clients? (10/100 or Gig-E)
How large are the clients?
What are your current settings?
Would you post the results from a Q OPT and a Q STAT?

The TSM Client Acceptor Daemon helps alleviate memory issues with the TSM scheduler (it's never been great at releasing memory and tends to hang periodically) by starting and stopping the scheduler as needed. It wont necessarily help you with the CPU usage of the TSM server.

 

[BBB]

On top of checking out your environment and tuning settings like Chad has suggested-

Are you using prompted or polling scheduling?

If prompted mode, setting up several schedules and splitting your nodes between them will help reduce your peak cpu usage. Just setup the schedules however you like, and give them descriptive names eg. DAILY_0300 DAILY_0400 etc.

If you're using polling mode, you could look at your schedule randomisation setting.

But even though this will reduce your peak load, checking your tuning and environment would be good too.