Don't underestimate the coupling of "simple" and "reliable".  Mainframes are
slow to change because they're not intended for change.  They're intended to
be reliable above all else.

The most reliable Web systems I've built had load-balanced sacrificial Web
servers (where reliablility was no big deal; just reboot if need be) tied to
a Tandem mainframe on the back end.  I think they took the Tandem down once
a year for preventive maintenance, but it was up otherwise.  This was a very
appropriate use of both very different kinds of technology and attitudes
towards change.

-glenn

> -----Original Message-----
> From: Mike Coleman [mailto:mkc@mathdogs.com]
> Sent: Tuesday, July 17, 2001 11:38 AM
> To: jim@itdepends.com
> Cc: 'kclug@kclug.org'
> Subject: Re: will lack of corporate support kill off Linux?
>
>
> Jim Herrmann <b3d@kc.rr.com> writes:
> > Spoken like a *nix hacker.  ;-)  No, Sprint is still, and will
> be for a very long
> > time, a mainframe shop.
>
> Well, what do *I* know?  ;-)
>
> (The impression I had, I formed at PCS; I didn't realize Sprint was so
> different.)
>
> > IMS fast path is still the very fastest data base in the world,
> and is used by
> > companies that need extremely high transaction volume.  Were
> talking tens of
> > thousands of transactions per second.
>
> I don't doubt this, but I wonder why they don't participate in the TPC
> benchmarks (see www.tpc.org).
>
> > Unix variants are pretty reliable, but not compared to the mainframe.
> > OS/390 has something like an average 99.9999% uptime.
>
> It'd be interested to track down exactly how and why this is (or
> isn't) true.
> In my limited experience in a industrial Unix environment,
> virtually all of
> the downtime seems to be caused by operator error, or, to a lesser degree,
> application errors.
>
> Just handwaving a bit, maybe 390's extreme reliability is due to the
> (possible) fact that very few new people are learning 390 and
> very little new
> code is being written for it.  If this were true, I suppose it
> would have the
> effect of significantly increasing the reliability of the
> platform (since the
> people would mostly be masters, and the code would mostly be debugged).
>
> > The mainframe is NOT dead.  It's merely been reinvented as a REALLY big
> > server.  :-)
>
> Yes.
>
> Hardware is hardware, I think.  If OS/390 is anywhere near as ugly and
> primitive as OS/400 though <donning my flame suit>, I'd much
> rather use Linux
> on mainframe hardware instead.
>
> --
> Mike Coleman, mkc@mathdogs.com
>   Windows XP
> http://www.mathdogs.com
>     Linux :)
> problem solving, expert software development