Installation option Z31=yes to force default to new design

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Hi,

We hadden some problems with abendU0046, msgice0046a and Ibm provide us with a fix (PK25047).

This fix introduce a new design. It is possible to switch from old to new design with installation option Z31=yes to force default to new design or with debug $HPHP0$ for individual jobs. It is also possible to force old design for individual job when in new design by default.

All that is nice, but i didn't manage to found description of this new design or what will be the effect of the z31 on my machine (performance, way to work or what ever) .

Could anybody provide me with some explonation or a pointer to some documentation ?

Many thanks beforehands,
Michel

William Thompson · Posted: Fri Jun 08, 2007 4:06 pm

I'm no dummy, but I have no idea of what you are refering to......
Anybody got a pointer to PK25047?

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Hi,
Here follow adresse of the documentation for PTF PK25047 at IBM site
www-1.ibm.com/support/docview.wss?rs=0&q1=uk24705&uid=isg1PK25047&loc=en_US&cs=utf-8&cc=us&lang=en

Ptf introduce a new design for dfsort with new option in dfsort allowing better utilisation of Hiper space in dfsort. The question is not for a bug or an abend but on how to use these news features of dfsort.

Best regards,
Michel.

William Thompson · Posted: Fri Jun 08, 2007 4:46 pm

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Hi,

That's just my question. But attention, we speak about Dfsort and not mvs It is a "design" in the way to dfsort works. The Z31 and other $HPHP0$ are Dfsort options, nothing to see with mvs. These options must modify the way dfsort manage its working storage (hiper space). That's why i placed this post in dfsort first.

Best regards,
Michel.

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Sorry i forgot to see that the new design is implemented by the ptf so, of course it is implemented if you install the ptf.

William Thompson · Posted: Fri Jun 08, 2007 5:06 pm

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Z31 is a new installation option introduced by ptf UK24705. ptf date is 07123 (03 may 2007). I didn't manage to found information about this new parm, that's why i open this post.

Best regards, Michel.

Frank Yaeger · Posted: Sat Jun 09, 2007 9:03 pm

Frank Yaeger · Posted: Mon Jun 11, 2007 9:27 pm

The guy I have to talk to about this is on vacation today. I'll respond after I talk to him.

Frank Yaeger · Posted: Wed Jun 13, 2007 11:04 pm

Here's the response from the DFSORT Team:

Prior to PK25047 DFSORT would allocate a portion of hiperspace it planned to use (based on available resources) at the beginning of the sort. When that initial Hiperspace was exhausted, DFSORT would allocate the additional hiperspace only if it determined the resources were still available.

After PK25047, when using Z31=YES, DFSORT will allocate all of the planned hiperspace at the beginning of the job. This will reduce the chance of an ICE046A being caused by DFSORT trying to get more hiperspace later in the job and it possibly not being available.

For similar workloads, the performance on your system will not be dramatically different before and after using Z31. DFSORT will still take into account available resources when determining the amount of Hiperspace to allocate. However, you may see some differences when comparing the same job, from those similar work loads before and after using Z31. Also, you may see some differences in the amount of hiperspace used by individual jobs. Some jobs may use more hiperspace and some jobs may use less hiperspace but you can be certain that DFSORT will use appropriate amount of total hiperspace, based on what is available on the system when the jobs run.

mgl · New User Joined: 12 Dec 2005 Posts: 29 Location: Belgium

Hi Frank,

Many thanks for your answer.
It sound clearer to me and i am happy with your answer. I think that i will reduce use of hyperspace (with hiprmax=p%) first in order to reserve enough memory to other job submitted, jobs here are submitted by block, not great but i didn't manage to let change job scheduling. Later, if no problem, try to leave hyperspace usage again at dfsort discretion (optimal).

Many thanks again for your help, have a great day,
Michel.