It's easy to diss a solution as 'budget' when it saves someone *else* money. 
The notion that a third CEC for standalone CF is substantially better than ICF 
is illusory. If you truly believe that one extra CEC is necessary, then you 
really need two extra CECs for CF because there are times when you have to take 
one of them down. Maybe for repair, maybe for upgrade. So you still need a 
backup. 

OTOH, 'logical standalone'--interesting term--with an ICF in each z/OS CEC is a 
sufficient solution. We've run this way for years and survived two separate 
hard-down CEC failures with zero recovery agony. My recommendation is to run 
two CECs, one substantial box to run application hosts, and a minimal 'penalty 
box' just for ICFs plus a few high-cost products that bill significantly less 
on a small CEC. With proper structure duplexing, you have extraordinary 
redundancy at a reasonable cost.     

.
.
J.O.Skip Robinson
Southern California Edison Company
Electric Dragon Team Paddler 
SHARE MVS Program Co-Manager
323-715-0595 Mobile
626-543-6132 Office ⇐=== NEW
[email protected]


-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf 
Of R.S.
Sent: Tuesday, July 10, 2018 3:00 AM
To: [email protected]
Subject: (External):Re: Sysplex between two hardware

W dniu 2018-07-10 o 06:56, Timothy Sipples pisze:
> I should also respond to this part:
>
> Radoslaw Skorupka wrote:
>> ...for availability reasons one should avoid having CF and z/OS LPAR 
>> on the same hardware, which means....
> That's not phrased as IBM would phase it, and it's not correct as written.
>
> Even when there's some merit in physically separating the CF, the 
> physical separation need only be between that CF and the particular 
> z/OS Parallel Sysplex it serves. Having other z/OS LPARs, even LPARs 
> that are participating in other Parallel Sysplexes, on the same 
> machine as the CF is consistent with IBM's recommendations here.
>
> David Raften has published a whitepaper (updated in May, 2018) that 
> explains the various CF configuration options. The direct link to the 
> current edition is here:
>
> https://public.dhe.ibm.com/common/ssi/ecm/zs/en/zsw01971usen/zsw01971u
> sen-.pdf
[...]

I quickly browsed the document and found confirmation what I learned several 
years ago.
That means:
Yes, you need either standalone CF (*see note) or CF structure duplexing! This 
is required for *some* structures, but ...it is. David Rften call it structures 
which /require failure independence/.

Note: Standalone CF should be understood here as LOGICAL stand alone CF.
IBM-MIAN do not allo pictures, so I'll try to describe it:
SYSPLEX PROD:
z/OS 1 in CPC A, z/OS 2 in CPC B, and CF2 in CPC B or A. And CF1 in CPC C.
Three CPCs.
However CPC C may be used for other purpose, i.e. for test LPARs. 
Anything but z/OS image belonging to the sysplex PROD.

There is also two-CPC configuration
z/OS 1 + CF1 in CPC A, z/OS 2 + CF2 in CPC B and *some structures are duplexed*.
What I heard unofficially from IBM-ers is such configuration is "budget" 
read: when you cannot afford good one.
And the link between CFs should be really local (short).


Note2: it is also possible to build a configuration without dedicated 
(logical stand alone) CF and without duplexing. Your choice.
However it is NOT resistant to all possible (single) failures. Yes, it 
is still better than parallel sysplex with single CF, but still 
vulnerable to some failures.

-- 
Radoslaw Skorupka
Lodz, Poland


----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to