Phil;

This is a great comment.  I hadn't given that much thought to the question.

Not to split hairs, but I didn't say MIPS, I said hardware.

If I had to guess, MIPS usage might actually increase slightly, because the
Pipes dispatcher has to switch between stages twice for every record that
is passed.  Access method overheard would drop.  Buffered access methods,
in most cases, only have to increment the pointer into the block buffer,
check for end-of-buffer and return to the caller.  I don't know for sure
which is larger.  Maybe someone more knowledgeable than I can shed some
light.

I would say the real savings would be in elapsed run time and working set
size.  Run time, due to eliminating something like 80-95% of I/O operations
for intra-JOB datasets.  Working set reduction would save on real memory.
(See below.)  Run time is probably more of a concern to customers,
especially those with tight batch windows.  That said, working set size
reduction would mean that processors would likely spend more, if not all,
time pegged at 100% busy, because so many more address spaces (TSO and JOB)
would be in a swapped-in and ready state than before.  Depending on what
metrics the capacity planners are looking at, CPU sales might actually
increase.  As I think about it more, if thru-put increases, new data could
be generated more quickly and other times of hardware could be more in
demand during peak load times.  I just don't know enough to say for sure.

Phil and others know what follows.

For those who don't know, in the typical case, a record passes through all
possible stages before the next record begins the same trip.  Each record
stays in the working page set, at least partially, during the entire time.
Parts that are referenced have a good chance of staying cache resident
between stages.

Think of it this way:  You can visualize UNIX piping as a series of
hourglasses open at both ends and connected in a tower.  Each grain of sand
must stop at every "stage" and wait its turn to go through the narrow
opening at the waist of each hourglass.  In Pipes, most stages have no
delay and it's like a single tall hourglass tube with only one narrow
point.  My best guess is that Pipes, in this analogy, would have only
5%-15% of the narrow openings as an equivalent UNIX piping command, meaning
that the data (sand) would flow 85-95% faster in the Pipes "hourglass".


OREXXMan
Would you rather pass data in move mode (*nix piping) or locate mode
(Pipes) or via disk (JCL)?  Why do you think you rarely see *nix commands
with more than a dozen filters, while Pipelines specifications are commonly
over 100s of stages, and 1000s of stages are not uncommon.
IBM has been looking for an HLL for program products; REXX is that language.


On Mon, Sep 20, 2021 at 12:48 PM Phil Smith III <[email protected]> wrote:

> Hobart Spitz wrote, in part:
>
> >The case *for *Pipes in the z/OS base.:
>
> > 2. Hardware usage would drop for customers.
>
>
>
> From IBM's perspective, that might not be a positive argument. It should
> be-they're hopefully not fooling themselves that they have a lock on
> enterprise computing any more, so anything that makes life more palatable
> for the remaining faithful at low cost to IBM should be A Good Thing-but I
> can easily imagine someone saying, "We estimate this will reduce MIPS
> purchases by x%, that's bad, don't do it".
>
>
>
> Just sayin'.
>
>
>
> ...phsiii
>
>
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO IBM-MAIN
>

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

Reply via email to