Thanks, Barton. 

The spool is pretty volatile. We may go from 30% to 95% and back down to
30% during the course of a day, sometimes 2 or 3 times. It doesn't take
very many catastrophic dumps from TPF guests to dramatically affect the
numbers. Spool bounces around quite a bit.

I am not talking Penguinese here. The guests are TPF. How much external
storage for a 5G or a 10G TPF system? There is no V-disk requirement, so
I would imagine that the per-machine penalty would be no more than the
virtual memory commitment, perhaps less.

I totaled the virtual memory sizes of the guests at the time of the
warning about PAGE being 100%. The result was 178G on a system that had
46G main and 10G XSTORE, and about 76G in page dasd. Even with the
paging to spool, the max spool reached during that period was 35%. We
only have 9 spool packs, so the total available spool space was much
less than the difference of 46G (178 - 76 - 56)

We are only going to see virtual storage demand increase as z/TPF
becomes a reality, so figuring out the size of the paging farm required
to support it will become important.



Regards, 
Richard Schuh 

-----Original Message-----
From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On
Behalf Of barton
Sent: Thursday, July 19, 2007 4:51 PM
To: [email protected]
Subject: Re: Address Spaces

Address spaces:  ESAMON ESAASPC
Page space, if you add a 4GB linux server, you need to add 8GB page
space.
paging in Spool space: look at ESAMON ESAPAGE prior to paging to spool,
gives you the 
spool requirement unless your spool is very dynamic.



Schuh, Richard wrote:

> Is there a command that will list all non VMDBK and non V-disk defined
> address spaces? Surely, there must be, but I am going blind looking
for
> it. For example, we have several dataspace enabled DIRC directories. I
> want to see which of them are currently defined, and any relevant
> details (size, number of users, etc.) about them, without keeping a
list
> of possibilities and querying each of them.
> 
> Somewhat related - is there any good way to predict the use of paging
> space? Normally, we run with <20% up to 40% of page allocation in use.
> We spiked today and occupied 100% plus whatever spool was required. At
> the time, we had 19 3390-03 equivalents in our paging farm. I looked,
> and less than 5% was devoted to V-disk. I don't know about the other
> possible defined spaces. 
> 
> Also, is there any display that shows how much of the spool space is
> being used for paging?
> 
> Regards, 
> Richard Schuh 
> 
> 
> 

Reply via email to