Kees,

It all helps and it's always nice to know others are doing it successfully.
Your comments on SCMFSD are particularly interesting (not at all you say) as
I'm sure some cleaning up is possible there.

I read somewhere that IBM expect nearly all requests to become asynchronous
eventually.  Processors are becoming quicker whereas the speed of light
isn't, so the heuristic algorithm used will deem spinning too costly for
shorter and shorter distances.

Groetjes,
Alan


-----Original Message-----
From: Vernooij, Kees (ITOPT1) - KLM [mailto:[email protected]] 
Sent: 11 January 2018 11:03 am
Subject: Re: SYSPLEX distance

If this helps: 
We run a parallel sysplex with sites at 16 - 18 km (2 separate routes with
some difference in distance) with active systems and CFs at both sites,
without problems.
Most Sync CF Requests to the Remote CFs are converted to Async.
To minimize the Async/Remote CF delays, we configure structures over the CFs
in such a way that the most busy or most important structures are in the
busiest or the most important site.
We do not use System Managed Coupling Facility Structure Duplexing. All our
applications are able to recover their structures well.
SMCFSD's inter-CF communication would add a number of elongated delays to
each CF update request. The advantage of SMCFSD is that each site has a copy
of the structure and intelligence can chose the nearest (=fastest) CF for
read requests.

Kees.

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

Reply via email to