Re: Article on new mainframe use

On Thu, 15 Aug 2024 16:04:13 +0000, George Neuner wrote:

On Wed, 14 Aug 2024 20:14:57 +0000, mitchalsup@aol.com (MitchAlsup1)
wrote:
>

It says:: "please enable JS and disable add-blocker"
>
ha

>
8-)  I do allow JS, but I tightly control it with NoScript.
>
>
>
The article started ok, talking about how certain applications are
latency sensitive and how they could benefit from having an AI be
resident on the mainframe.

There are only 3 kinds of applications::
a) those that are latency sensitive
b) those that are bandwidth sensitive
c) those that are balanced between latency and bandwidth.
Caches deal with (a)
Vectors and/or GPUs deal with (b)
(c) occurs few and far between.
All SPEC benchmarks are (a)

Actually a good portion of the article seemed rather positive on the
idea of implementing AI models on mainframes.  Unfortunately it
devolves into implications that AI doesn't play nice with COBOL and
that mainframes can't be programmed in other languages.

The only way AI would not play nice with COBOL is if COBOL has no
data type appropriate to the AI interface(s) as an argument or as
a result. {{You can argue if this is a COBOL problem or an AI
problem}}

Not knowing COBOL well, I can't speak to the former.  The latter,
however, is bullshit.
>
>
from the article:
"
Mainframes do have limitations. They are constrained by the computing
power within their boxes, unlike the cloud, which can scale up by
drawing on computing power distributed across many locations and
servers. They are also unwieldy—with years of old code tacked on—and
don’t integrate well with new applications. That makes them costly to
manage and difficult to use as a platform for developing new
applications.

In 1982 I was sitting a room of sever TI silent 700 keyboards, logged
onto a CDC 7600 in San Diego typing in 8085 ASM code; when all of a
sudden the room went quiet (so much for the Silent part of Silent 700)
About a minute later the clatter began again, this time I was still
logged onto a CDC 7600, but not it was in Chicago, and I had all my
files there ready for work. Apparently the San Diego machine took a
hard crash, and the PPs shipped the workload to Chicago where we
continues to work as if nothing had happened....
1982 people--NOS--pre internet...

>
   :
>
Still, efforts to do away with the mainframe have been going for
years, especially as the number of developers conversant in Cobol—one
of the primary programming languages used in mainframes—quickly
dwindles.
"

While dusty deck COBOL programs make assumptions about the environment
they compute within, there is nothing preventing computer makers from
supporting those features that COBOL programs require and getting a
COBOL compiler for their ISA up and running. The problem is that as
long as there are mainframe makers, the other computer makers do not
see a big enough market to make the required investments.