>> On Mon, 2 Feb 2009 13:25:06 +0100, Christian Svensson <Christian.Svensson AT 
>> CRISTIE DOT SE> said:

> The issue I have is that the customer only want to runt TSM on
> Windows and not AIX or even Linux.  But to virtualize the TSM
> Servers we maybe can have more TSM Servers on lest Intel
> Hardware. This is just a test and see what we can do.

Well, that's interesting.


OK, so to restate:

When your windows installations bottleneck, the _cause_ is high IO,
but they _symptom_ is 100% cpu utilization.  You suspect some sort of
thrashing is going on, so you're hoping that by splitting the peak IO
into separate streams you'll be able to keep the CPU load under
control.

I still think that the case for virtualization is very weak
here. You've got enough work to occupy an entire physical box with one
system image.  This makes me ask "What are you getting out of it?"

If the customer is completely running on virtual infrastructure to
begin with, it may be more pain to go physical than it's worth.  Local
work habits are certainly a good reason to pick the virtual path.

But TSM would be the last think I would virtualize, not the first, or
even the fifth.


I'm a little astonished that the load you describe can flatten the
hardware you describe.  Is there some other way in which your
environment is exceptional?


- Allen S. Rout