Groups | Search | Server Info | Login | Register

Groups > comp.os.linux.misc > #35834

Re: COBOL and tricks

Cross-posted to 2 groups.

Show all headers | View raw

Bob Eager <news0009@eager.cx> writes:
> I did some work for a large financial company, writing a compiler for a 
> query language that had been compiled and interpreted by COBOL programs.
>
> I got a BIG speedup, enough to secure some new contracts.
>
> They (the UK office) then asked me to implement it on another kind of 
> machine. We discovered that the US parent office (in Boston, I think) had 
> declared that the Brits were wasting their time and it couldn't be done.
>
> The UK guy had faith in me. And we got a massive speedup there too.

turn of the century, large financial outsourcing company that did half
of all issuing (i.e. credit cards) in the US (embossing,
statmenting/billing, transactions, call centers, etc) ... had 40+ max
configured IBM mainframes (@$30M) (number needed to finishing settlement
in the overnight window), all running 450k cobol statement program
... they did have large group that had been responsible for the
performance care&feeding for decades ... but had gotten somewhat
myopically focused on instruction hot-spots.

I was to look at throughput ... using some performance analysis
technology from the early 70s at the ibm cambridge science center and
found 14% improvement. Another person was brought in using different
technology that found another 7% (total 21% throughput improvement for
>40*$30M operation).

What was funny was the other technology, he had aquired in the early 90s
(during IBM troubles and unloading lots of stuff) an APL system analysis
program that he then ran through an APL->C converter ... he had been
using it for (both IBM and non-IBM) datacenter performance consulting
business (around the world)

It turns out it was a descendent of the IBM HONE (world-wide online
sales&marketing support) 70s "Performance Predictor" (IBM reps could
enter customer workload&configuration questions and ask "what-if"
questions regarding changes to workload &/or configuration) ... which
was an APL-based analytical system model (also) developed at the IBM
Cambridge Science Center.

trivia: mid-70s, US HONE datacenters were all consolidated in silicon
valley and enhanced to have eight mainframe systems configured in
loosely-coupled, single-system-image with load-balancing and fall-over
across the complex (load-balancing decisions were implemented with a
modified version of the "performance predictor").

-- 
virtualization experience starting Jan1968, online at home since Mar1970

Back to comp.os.linux.misc | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

Re: COBOL and tricks Charles Richmond <codescott@aquaporin4.com> - 2022-10-28 02:37 -0500
  Re: COBOL and tricks Charlie Gibbs <cgibbs@kltpzyxm.invalid> - 2022-10-28 22:16 +0000
    Re: COBOL and tricks Vir Campestris <vir.campestris@invalid.invalid> - 2022-10-30 20:47 +0000
      Re: COBOL and tricks Bob Eager <news0009@eager.cx> - 2022-10-30 22:18 +0000
        Re: COBOL and tricks Anne & Lynn Wheeler <lynn@garlic.com> - 2022-10-30 14:20 -1000
      Re: COBOL and tricks scott@slp53.sl.home (Scott Lurndal) - 2022-10-31 14:43 +0000
        Re: COBOL and tricks Peter Flass <peter_flass@yahoo.com> - 2022-10-31 09:37 -0700

csiph-web