On Thu, Apr 30, 2009 at 6:21 PM, Rich Smrcina <rsmrc...@wi.rr.com> wrote:

> Both ESALPS and the Performance Toolkit will show that an unused Vdisk will 
> use very little storage.  Even if you do manage to start using it, after a 
> certain amount of time it is likely to get paged out if unreferenced long 
> enough.  Barton has been presenting these topics at SHARE and WAVV for 
> several years now, they should be well burned in...

I can't speak for the other product, but you could be right ;-)

The reason we tell you not to fill that second overflow VDISK is not
because it would take up too much real memory, but because it will
make things very slow even though you did not want it to. And because
when VDISK gets paged out, you need twice the amount in page space.

The overflow swap VDISK is like what normal swap is for Linux on other
platforms: it's supposed not to get used unless the application went
out of its mind or something like that. Assuming that not all servers
go crazy at the same time, you can probably stand it reasonably well
if you set up monitoring.
When the workload slightly grew beyond expectation, it will drip into
the 2nd VDISK only a little. That means it will take a while before it
really takes up space, and you have time to handle it. Even when
multiple servers seem to outgrow plan. As long as it is VDISK, people
will not notice and you can plan for change. But when your 2nd VDISK
is real disk, performance will be bad (even with only a little
swapping) and big badges will prevent easy planning of the change.

Rob
-- 
Rob van der Heij
Velocity Software
http://www.velocitysoftware.com/
  • ... Martin, Terry R. (LOCKHEED MARTIN Performance Engineering/CTR) (CTR)
    • ... Mark Post
    • ... Adam Thornton
    • ... Barton Robinson
      • ... Scott Rohling
        • ... Rich Smrcina
          • ... Scott Rohling
            • ... Mike Walter
              • ... Rich Smrcina
                • ... Rob van der Heij
            • ... Martin, Terry R. (LOCKHEED MARTIN Performance Engineering/CTR) (CTR)
              • ... Marcy Cortes
                • ... Martin, Terry R. (LOCKHEED MARTIN Performance Engineering/CTR) (CTR)
        • ... Doug Shupe
          • ... Barton Robinson
          • ... Tom Duerbusch
      • ... Martin, Terry R. (LOCKHEED MARTIN Performance Engineering/CTR) (CTR)
        • ... Barton Robinson
          • ... Martin, Terry R. (LOCKHEED MARTIN Performance Engineering/CTR) (CTR)
    • ... Phil Smith III

Reply via email to