Re: IBM: Linux is ready for the Mainframe. SUN: No it's not!

From: Timothy J. Bogart (tbogart_at_frii.net)
Date: 08/10/04 

Date: Tue, 10 Aug 2004 12:52:59 -0600

Michael Kraemer wrote:
>>Thanks for the data point. I have used Applix purchased in about 1990
>>for AIX 3.1 just fine on 5.2,
>
>
> How comes ? If I remember correctly their was a major break
> between AIX 3.1 and 3.2. Many, if not all, apps needed to be
> recompiled. This was in the very early days, shortly after AIX was
> introduced.
>
>
>>but my knowledge of the old Sun changeover
>>was unfortunately second hand horror stories. Normally a good source,
>>but perhaps this time he was skewed to the problems. I was left with
>>the impression everything was hosed.
>
>

Well, it was a long time ago, but I remember we had a 520 when it was a
hot new machine, and the xstation 120 with documentation marked as
'preliminary', and I just don't remember any such break. Course, at the
time, I wasn't looking for one either.

If anything, you might criticize the 1.x and 2.x for how short lived
(relatively speaking) where there was no binary compatibility moving
across different hardware.

Then again, a memory of a sparcstation based on x86 (classic?) creeps in
the background - not going to have binary compatibility either.

Wait one, perhaps you are thinking of the great common-mode blunder?
Somewhere in the introduction of one of the PowerPC chips, they didn't
implement all the instructions of the POWER architecture, so you had
some issues of code that would try to use those instructions and go
boom. Hence, the options were introduced to compile optimized code for
one family or another, or code guaranteed common to all processors.

I am not sure how much code was affected, but I think we ran into one
instance of commercial code (Wordperfect, I think). I do know I gave my
IBM friends crap about it for years afterward.

One yet another hand, there was a time when I was asked to pull together
some ancient hardware to run that Netscape web server (drawing a blank
on the name) and after some hours of banging away with no success, the
support folks told me it wouldn't work on that old hardware. Needed the
next release CPU. Sure enough, an newer machine and the install and
setup worked like a champ.

Contrast that with a similar situation for an Oracle Enterprise Edition
for a small development group. I was able to meet the minimum
requirements for that beast with an pulled-from-the-trash 250 (as I
recall, the first PowerPC implementation) and the development group was
able to test against an instance of exactly what was going to be used in
production - and not a similar (smaller) version that should work the
same. Just what the folks wanted to do with that web server on the Sun.

Is there a point to all this? I hope so. We can have fun dredging up
our old memories of how well things have worked - and it is especially
valuable to swap war stories when we fall into the trap of thinking our
favorite vendor is perfect - there just ain't no such thing. And it is
not like we have the equivalent of a Consumer Reports to go look up
serious data on reliability. We are stuck with our own experiences, and
perhaps experiences of some folks we learn to trust (and this portion of
the thread points out even that is not quite perfect, eh?).

Sort of makes one wish there were an 'open-source' project to collect
actual documented problems/gotchas and fixes for all the different
vendors so you could make a serious comparison.

As long as we are swapping stories: one of the reasons I generally
prefer IBM is that I have had personal experience with how to make them
sit up and take notice when they make a serious mistake - and seen them
come thru. Back in the early J50 days (we even had J40s upgraded to
J50s - amazing the hardware hoops you get to jump thru to get around
crappy software) I was getting pushed out of sysadmin into
sysengineering where the new J50s were being put thru their paces to
verify performance. After my first week, I got on the phone with our
interface with IBM and yelled 'What the F--- is with IBM? I thought you
folks sold solid hardware. Anything you care to tell me about the J50s
you didn't mention when you sold them to us?' To this day, that is the
only time I can recall using the real F word in such a situation. And,
to her great credit, she knew I wouldn't do that unless something
serious was going on. "I don't know, but I will find out" was the very
appropriate reply. Within an hour, I had it confirmed there was indeed
an engineering issue with the CPU cards, allegedly affecting 20% of J50
customers who exercised (exorcised??) the hardware in a certain way.
There was an engineering fix being tested, and I was assigned an
engineer in Austin who was required to be in touch with me at least once
per business day until the problem was solved to our satisfaction.

Now, for the most part, all the major vendors must do this at some
point, but it sort of sticks when you have a serious problem and they
come thru for you in a crunch.

These flame-bait sort of things can be useful when actual data points
are shared. Quite distinctly different from some of the 'well, these
folks bought Windows, so it must be good' sort of crap you sometimes get
(speaking of flame bait). 8-)

Relevant Pages

  • Re: IBM: Linux is ready for the Mainframe. SUN: No its not!
    ... across different hardware. ... Just what the folks wanted to do with that web server on the Sun. ... Sort of makes one wish there were an 'open-source' project to collect ... J50s - amazing the hardware hoops you get to jump thru to get around ...
    (comp.unix.aix)
  • Re: Recovery through motorbike vacation
    ... that went bad and needed a hardware swap. ... I've always wondered what folks on the same HV distribution feed as a large ... turn the large arc furnace on and off. ... But stuff does get led out to the w0rng place. ...
    (alt.sysadmin.recovery)
  • Re: Conroe L1 cache ??
    ... change from their previous generous policy of access to detailed hardware ... the irony is that most folks don't really read them. ... sheets get less data on them...especially as the folks buying the CPUs ... Intel needs the tuning documents for their customers ...
    (comp.sys.ibm.pc.hardware.chips)
  • Re: The Garmin 496...a teenagers review
    ... the software that I used was a dog. ... I hear folks ... blaming the hardware and the .NET framework continually when it really is a ...
    (rec.aviation.piloting)