BatchPipes was useless for any really large volume of data.  My employer had a 
license to the product early in my time with them (OS390 era) and we looked at 
using it for efficient data passing across multiple batch stream jobs, but our 
very large data volumes precluded using it.

Also, BatchPipes was not really a PIPEs implementation – it was just a way to 
push data from one batch process to multiple other ones without a real file as 
the intermediary, the idea being that elapsed time could be reduced thus 
improving SLA compliance.

When I think of PIPEs I mean a CMS-style pipe command implementation usable in 
a single (possibly) multitasking process, not BatchPipes sharing data across 
jobs.

Your obvious prejudice that business application teams don’t know what they are 
doing is really irritating.  We DO know the system, and we DO design to use the 
best-of-breed features that z/OS provides, and we DO consult with systems and 
capacity and storage and database and network teams to help us do our jobs 
better.

Systems-level teams are NOT the ultimate deciders of what is acceptable 
performance design, and z/OS is NOT, as you put it, “. . . designed for 
sysprogs making choices”.  Systems teams are PART of the business application 
implementation team.  If a true multitasking application process (using real 
PIPEs or not) made business sense from a business perspective, then 
consultations with systems-level teams would help support that implementation, 
not bad-mouth the application team for even thinking about such an 
implementation.  The job is to get the application running as best as we can to 
support the business effectively and efficiently, not to have arguments about 
which implementation technique is “better”.  Systems teams do NOT rule the z/OS 
universe – business applications do, because that’s what the business pays all 
of us to implement.

You say you are debunking the myth that PIPE’s are performant, but we’ve never 
seen a real and complete z/OS PIPEs implementation with a similar set of stages 
and filters as are available in CMS that would allow for a side-by-side 
comparison to prove or disprove the performance of an application using PIPEs 
versus other z/OS techniques.

All this is frivolous watercooler talk anyway, since z/OS doesn’t have a real 
PIPEs implementation (and likely never will have one) outside of Netview, which 
effectively means none at all since AFAIK business applications can’t use 
Netview for business application purposes (i.e., not in the context of 
automated message processing but for real business data processing).

Peter

From: IBM Mainframe Discussion List <[email protected]> On Behalf Of Jon 
Perryman
Sent: Monday, September 8, 2025 8:10 PM
To: [email protected]
Subject: Re: Pipelines = you don't understand z/OS

On Mon, 8 Sep 2025 06:45:45 +0000, Farley, Peter 
<mailto:[email protected]> wrote:

>A great deal of your arguments seem to me to be about the performance 
>impact of a multitasking program on other-system-critical applications, 

I'm discussing why TSO PIPEs no longer exist. I'm saying that people use them 
because they are perceived as performant. I'm debunking that myth. There are 
alternatives that could be made as performant that don't involve PIPEs.

>How is the performance of one truly multitasking batch step any different from 
>any other WLM-managed batch job? 

Batchpipes involved 2 or more jobs. Should IBM implement batch enclaves? Or 
maybe IBM should implement multi-tasking JCL definitions?

>You seem to be arguing that a multitasking batch job step in a production 
>batch stream could interfere 
>with the priority-CPU-I/O of the online and real-time and database activities 
>running on the same system(s).  

I'm saying that batchpipes most likely failed because it ignored performance. 
I'm saying that batchpipes removes control and understanding by the performance 
sysprog. 

>I just don’t understand why you are so dead set against pipes for "allowing" 
>multitasking 
>to be more easily created by "ordinary" programmers.  

I'm not against pipes, however PIPEs are designed for uncomplicated 
environments that don't have sysplex, srm, sysprogs and more. z/OS is designed 
for sysprogs making choices whereas Unix is about application programmers 
pretending to be in charge. 

How many times will Unix distributed computing be allowed to fail (now moving 
on from Kubernettes to micro services)? SYSPLEX has been around for at least 30 
years while Unix develops a new distributed solution every couple of years. 

Doing things right the first time is uniquely mainframe.
--

This message and any attachments are intended only for the use of the addressee 
and may contain information that is privileged and confidential. If the reader 
of the message is not the intended recipient or an authorized representative of 
the intended recipient, you are hereby notified that any dissemination of this 
communication is strictly prohibited. If you have received this communication 
in error, please notify us immediately by e-mail and delete the message and any 
attachments from your system.


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

Reply via email to