On Sun, 22 Sep 1991 08:07:05 EST Nick Laflamme <[log in to unmask]>
said:
>Raise your expectations and challenge IBM to meet them. IBM can't be all
>things to all people, but it can do more than you expect.
IBM won't fix anything major in SP5, because SP5 is being removed from
marketing very soon, so as to force customers to spend big money on ESA
systems. SP5 is what I am licensed for at the moment. SP5 is likely to be
what I am licensed for in 2 years. In 5 years I doubt there will still be
a VM system in the basement. So, there's no point in me trying to
challenge anything at the moment.
The reason is that (before I came to work here) we bought a rack 370 box.
Major mistake. To start with, the 4341 I was using in 1985 was not only
faster but above all more reliable. Marketing said this 9370 could
operate in a regular office environment, provided there was a reasonable
air conditioning system. Marketing also said the disk units were very
fast, 3Mb/sec transfer rate, 3380-level (anyone who uses 9335's knows
this is a good candidate for the Guinness Joke of the Year award, but
when you don't have one it's none too obvious). Well the facts are that
the box is in a professionally cooled computer room, that the disk units
decide, from time to time, to stop with a thermal check for no apparent
reason, that the CPU was down for about 24h and didn't come back until
all but one of the elements in the rack had been replaced, and that we
had to do the same for a disk unit. Even after being repaired, we have to
run it with cover open or it gives a thermal check (room temperature
around 20C/68F). The tape unit often decides to let go of the reel in the
middle of a backup. Several times a day, all terminals get stuck at the
same time for 15-30 seconds (clock icon). Once every other month, I have
to power cycle the whole system because the ethernet controller is all
dead. The more I think about it, the more it becomes clear that the only
part of the rack which works to satisfaction is the power unit.
For the cost of an apartment in the city (and they are not precisely
cheap), we can upgrade to a newer rack system, something that has ESA and
S/390 and all the newest whiz-bang. Still slow, but it can run ESA
programs. Or rather, it could, if it could run an ESA operating system.
You see, there is this little detail: rack systems come with FBA disks to
save a couple pennies and VM/ESA doesn't support these, so, yes, if you
boot from a tape and never want to do any disk I/O I guess you can run
ESA software. Of course, this box *can* run "VM/ESA" - the 370 feature,
the one that has nothing to do with ESA apart from the name. We get 80%
discount on that thing and the box is one of the smallest IBM makes, and
it's still the price of a good car.
Marketing says I shouldn't despair, FBA support will be back into VM/ESA
some day. Well I have written a bunch of code that had to dual-path for
FBA, and I know what that means. FBA support is not a one-day job, it's
going to cost IBM millions to put it back into VM/ESA. Before the
re-introduction of FBA devices in rack systems, only "ancient" disk units
(3310 and 3370) used FBA. IBM had a very strong case for saving that
money and saying FBA was a dead architecture, in spite of the fact that
it makes the logic cheaper to build. That is what the VM/XA developers
did, and I thought that meant the end of FBA and life would be simpler in
the future. But... They had to save on rack systems which have never and
probably never will be competitive, even with serious non-unix hardware
from DEC and others, so I guess we're going to drag that trash into VM
for another decade. And since most VM/ESA shops do not and probably never
will use FBA devices, most "system utilities" written by IBM customers
won't work on FBA systems, so the rack owners will be at a disadvantage -
all that because IBM didn't want to put real CKD logic on their rack
minidisks, to "save" money.
I realize none of that applies to mainframe computers. I used to be in
charge of a 3090, and they are excellent machines in all respects. The
only complaint I ever had about them is the time it takes to bring them
up after a scheduled power outage, due to the use of 4331's as support
processors. A 3380 isn't going to stop on a thermal check just because it
feels like having a bit of fun, not will an ancient 3420 ever release the
reel in the middle of a backup. An unaligned load on a 3090 doesn't take
45 times longer than an aligned one. Only a toy rack machine can have
such attributes. Unfortunately, IBM doesn't tell you about this when they
sell you a 9370. Do you know of any other computer manufacturer that
calmly tells you the cost of the license for the operating system of a 2
mips machine with 16M of memory is $85k + taxes (without academic
discount)? That doesn't even include any of the products you want to run
(multiply by 4 for TCP/IP, RSCS, C and PASCAL compilers plus a couple
other things you need on a small network machine like SEARN).
This is one of the reasons we're going to have to put up with a bunch of
people running "ancient" versions of the operating system for many years.
Another reason is the 4381 dead-end, but that's all going off the topic
of this list. The bottom line is that IBM still doesn't have any decent
offering for the smaller shops. The difference is that in 1985 it meant
you had to somehow find a whole lot of money to buy yourself a good
system, whereas now you have to find a lot of money to buy yourself an
unsatisfactory system, which will require another lot of money a couple
years later to keep running current application software. I'm afraid I
don't see this as an improvement.
Eric
|