On Wed, 5 Mar 2008 23:45:02 -0800, Walter Marguccio
<[EMAIL PROTECTED]> wrote:
>Scott,
>
>sorry for being persintent on this, but what would you then recommend for
RESMIL
>on a basic sysplex with 3 LPARs :
>
>a) 'hot potato' RESMIL=OFF (fast RSA on the ring, but CPU overhead)
>
>b) RESMIL=0 on one LPAR only (which one? The one with more resources?)
>
>c) RESMIL=0 on all LPARs
>
>I'd be glad if you could make a comment on this. Thanks in advance.
>
>Walter Marguccio
>z/OS Systems Programmer
>Munich - Germany
Without knowing a whole lot more about your system, I really can't make a
'recommendation'. What I can do is give you several principles that might help
you make a decision:
1) There is just a whole lot of important stuff in the system that relies on GRS
working quickly (Catalog, Allocation, Open, Close, etc. all of which are the
building blocks for getting user work done).
2) The longer you make important stuff wait, the more important stuff backs up
behind it.
3) The longer your things wait, the more opportunity there is to create
contention between them.
4) Response and or elapsed time are probably meaningful measurements for you.
5) Elongating and exacerbating 1, 2, and 3 negatively impact 4
So, if you accept those principles, then one would say that you should "Render
unto Caesar the things which are Caesars", that is, give GRS as much
horsepower as you can afford to.
It's clear that CPU utilization is a function of the 'total RESMIL' for the
complex
(since altering one impacts them all), so I'd spend a bit of time playing with
different values, plotting total RESMIL vs. total (well, really, XCF+GRS) CPU to
see how much you are willing to 'render'.
Something like:
Total RESMIL CPU
OFF OFF OFF c1
OFF OFF 0 c2
OFF 0 0 c3
0 c4
1 c5
2 c6
etc.
Scott Fagen
Enterprise Systems Management
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
