>From Jay Maynard:
On Fri, May 24, 2002 at 02:42:32PM -0700, Lionel Dyck wrote:
> > Seems that the linuxworld author of the mainframe articles is none too
> > happy with those on this listserv.
>
> I haven't been reading this guy's articles...and after reading the sidebar,
> it's clear I haven't missed anything. He simply Doesn't Get It.
If he got "it" he'd be able to point out what Mainframe Linux
is good for and why the lowered performance of the environment
on a mainframe is reasonable.
Perhaps he needs to read Appendix "A" of the Linux for S/390
Redbook- which compares the two architectures (Intel's Pentium
and the s/390) and points out that the s/390 isn't so much a
speed demon as it is a RELIABLE place to run _any_ workload.
Everything in the world is a trade-off. The s/390 architecture
trades some performance advantages of the technology for a big
boost in reliability- which would require a LOT of specialized
hardware to mimic on a Pentium-class chipset (i.e., a whole new
chipset, from ground up, would be needed).
Now I'm an AIX geek (wanna pronounce AIX? Here's a Hint:
AIX 'n Panes) and there are some things AIX does exceptionally
well (i.e. I/O) because it's internal architecture (which has
a vaguely Unix-like API wrapped around) doesn't look anything
like any Unix I/O system I've ever seen before (and I've seen
a few). Linux has a "conventional Unix" (if there is such a
thing) I/O architecture, so there may be some issues in getting
the most out of the S/390 architecture. Perhaps IBM should
bring back AIX on the mainframe but make it like AIX 5L, with
full Linux affinity.
Of course, with the new Power4 CPUs and their ability to handle
LPARs, perhaps it's only a matter of time before p/VM (for the
pSeries) gets put together. p/VM and CMS so that you can set
up the environment before booting the actual OS in a pSeries
virtual machine...
Consider the bane of computing: University Computing Centers,
where all those clever students are trying to "crack security"
(well, there were plenty of these kinds of people 30 years ago
at SUNY Stony Brook that I kept bumping into) trying to get
"something for nothing" out of the batch system (370-155) and
the timesharing system (PDP-10); I suspect the VM+Linux
environment looks like a godsend since a student can be issued
his own system w/ a root ID- and the VM operator can still
adjust the instance's priority and even crash it w/o having
to walk up to a physical machine. Additionally, such systems
should not be compute bound, so there's no huge waste of
potential cycles (the hardware can be amortized by other jobs).
I think it's this ability to balance workload to keep the CPU
busy that's most likely to explain this. IBM itself (we got
a presentation at the SunCoast LUG meeting) indicates that
existing compute-bound applications are "not a good fit"-
though web-hosting and file serving are two good fits since
they're bursty and don't sit and spin (unless the web host is
facing the "slashdot effect").
--
John R. Campbell Speaker to Machines [EMAIL PROTECTED]
- As a SysAdmin, yes, I CAN read your e-mail, but I DON'T get that bored!
Disclaimer: All opinions expressed above are those of John R. Campbell
alone and are seriously unlikely to reflect the opinions of
his employer(s) or lackeys thereof. Anyone who says
differently is itching for a fight!
