On 6/13/06, Larry Macioce <[EMAIL PROTECTED]> wrote:

Thanks Jim that got me on. But now I'm a little puzzled(as always) When I
started to view what files were in monwrite the file that had captured
data had 16171 rec at 4k apiece. With a little math that came up to almos
78 cyls. I'm 12 short. The other 3 file have 14 4k recs.

Your Math May Vary :-)  My math suggests that 16171 records takes 89.8
cylinders (at 180 records per cylinder). And then you need 16 or 17
pointer blocks for that one, some for directory and allocation map.
All together pretty close to 90 cylinders.

If you need to collect raw data it will be to have someone analyze
your performance problems. In that case they will need detail and it
would not be very useful to disable monitor domains until 90 cylinders
will last. Be prepared to set aside a full 3390-3 for raw data and
offload that several times per day. And since you cannot get rid of a
full pack well within a minute, an decent approach is to set up a pair
of full packs and have it switch between them (while another server
will process the data in the meantime). Indeed, not pretty and rather
expensive (but nothing compared to VMPRF processing that huge amount
of data).

As an alternative, our ESAWRITE server will process the *MONITOR data
in real time and keep compacted data in the performance database for
analysis and reporting. This is typically an order of magnitude less
space (and CPU cycles). In that case you would only need to capture
raw data when there is a bug in the VM monitor or to have IBM look at
your data.

Rob
--
Rob van der Heij
Velocity Software, Inc
http://velocitysoftware.com/

Reply via email to