VMS for those realy big critical applications
JF Mezei
why one of the largest and most critical systems, the airline reservation
system SABRE is switching to UNIX instead of going to VMS.
Does anyone know who SABRE chose as supplier for their main reservation system ?
If Galaxies is such a great advance and is unmatched in providing exactly what
SABRE needs for its mission critical systems, why would it have lost such a
sale ?
If Palmer has already restrained VMS to that niche, the SABRE reservation
system would have been exactly the ONE customer Digital should have pulled all
its tricks for to ensure it won that contract and would have given VMS and
Digital the visibility it badly needs for the market it has
restricted VMS to.
And ppppplease, stop using the Crédit Lyonnais as an example. It is old. Is
that really the only one that they can use ?
Wayne W. Scott
>
> If VMS is so robust and scalable and UNIX is unreliable, I am puzzled as to
> why one of the largest and most critical systems, the airline reservation
> system SABRE is switching to UNIX instead of going to VMS.
What is it switching from? Probably a big IBM mainframe!
Which version and platform is it being converted to?
Since AMR owns it and is in Texas, and IBM's main AIX & RS/6000
team is in Texas, I doubt they're thinking of Solaris on Sparc.
> Does anyone know who SABRE chose as supplier for their main reservation system ?
Not yet, I don't; but I used to work with a guy on a big AIX project in NJ.
He moved to Texas to work on Sabre. Hmmm, might be time to drop him a line.
> If Galaxies is such a great advance and is unmatched in providing exactly what
> SABRE needs for its mission critical systems, why would it have lost such a
> sale ?
Because it wasn't there when AIX and RS/6000 was.
Because IBM does aggressive marketing & Digital did stealth marketing.
Because Compaq, in Texas, hadn't bought Digital until it was too late
to go after another Texas customer.
> If Palmer has already restrained VMS to that niche, the SABRE reservation
> system would have been exactly the ONE customer Digital should have pulled all
> its tricks for to ensure it won that contract and would have given VMS and
> Digital the visibility it badly needs for the market it has
> restricted VMS to.
Yeah, but I doubt the sales force was very motivated during the fire sales.
Just my WAGs (Wild Ass Guesses).
Wayne
Jerry Leslie
: JF Mezei wrote:
: >
: > If VMS is so robust and scalable and UNIX is unreliable, I am puzzled
: > as to why one of the largest and most critical systems, the airline
: > reservation system SABRE is switching to UNIX instead of going to VMS.
: What is it switching from? Probably a big IBM mainframe!
Yes, here's a story about an outage of their IBM DASD back in June:
http://www.zdnet.com/zdnn/stories/zdnn_display/0,3440,2115078,00.html
The Sabre Group's home page is:
--Jerry,
Jerry R. Leslie jerry....@aspentech.com Aspen Technology, Inc.
(my opinions are strictly my own)
Jerry Leslie
: http://www.zdnet.com/zdnn/stories/zdnn_display/0,3440,2115078,00.html
: --Jerry,
Sorry to post a follow-up to my own post, but it doesn't look like
the Sabre Group is giving up their IBM big iron, and they do use
some VAX/VMS systems.
From Terry Jones, CIO of the Sabre Group, per:
http://www.sabre.com/news/tech_news.htm
"The question I get more often (although it is phrased as the one above)
is really getting at why is the SABRE system back-end mainframe-based
and not based on some UNIX or other architecture. This answer is simple,
there is simply no other processor complex, nor operating system with
the power to run the 5,000+ transactions per second that the SABRE system
requires. In addition, there is no other operating system that operates
at the 99.95 percent uptime that we require.
We aren't the only ones to make that conclusion. Major banks and most
credit card companies run their card processing on the mainframe based
TPF operating system for the same reason.
That is not to say that we aren't continually testing alternatives. We
have and we are doing just that. In fact many of the transactions that
the SABRE system processes will be eliminated over time as smarter
clients eliminate the need to position the cursor of remote terminals
or acknowledge security messages of ticket printers.
The SABRE system is an evolving system where a matrix of processors and
operating systems each play a role. Our direct connect products run on
Stratus computers under VOS, our total access products run on VAX computers
under VMS, our Cargo Routing Guide runs on Power Parallel under UNIX,
our Fares are prepared in MVS and executed on air-cooled CMOS machines,
and so on.
Maybe Mr. Gates and his 'billion transaction' machine can help us evolve
even faster...only time will tell."
JF Mezei
> > If Galaxies is such a great advance and is unmatched in providing exactly what
> > SABRE needs for its mission critical systems, why would it have lost such a
> > sale ?
>
> Because it wasn't there when AIX and RS/6000 was.
> Because IBM does aggressive marketing & Digital did stealth marketing.
> Because Compaq, in Texas, hadn't bought Digital until it was too late
> to go after another Texas customer.
Palmer and Compaq had been planning the takeover for 3 years.
Compaq has made attempts at getting DEC and failed and turned to Tandem
instead (before coming back to DEC).
Galaxies must have been in the works for a couple of years right ?
Now, considering the size and importance of such a customer, I would think
that Palmer would have told AMR about all those plans to hand over Digital to
Compaq and also allow VMS engineering to release Galaxies.
I get the feeling that VMS was not even in the running. Not sure if Digital
even was.
SABRE is one of the very few applications that *really* need all that fancy
Oracle performance, very large databases and very high transaction/second
capabilities. SABRE handles orders of magnitudes more transactions per second
than any bank does.
SABRE's data centre is centralised and is so critical that it is litterally
burried a few stories below ground in a bunker. Bank data centres are nowhere
nearly as protected as SABRE's.
JF Mezei
>The heart of our technological infrastructure is in Tulsa, Oklahoma, the home
to our
>120,000 square-foot high-security data center. Housed in the data center are
17 mainframes
>with more than 15.3 terabytes of electronic storage, over 4,000 MIPS of
processing power,
>180 communications processors and numerous midrange, UNIX-based computers.
This data
>center runs the SABRE Computer Reservation System (CRS) in a mainframe operating
>environment, the ideal solution for the high-volume, high-availability
requirements of the
>distribution business.
And yes, it seems those who said SABRE and IBM were close, they seem to be
right. Some news releases do indicate that IBM got quite a few contracts from SABRE.
It is a shame that DEC couldn't have flaunted its technology and replaces IBM.
Now the really serious question: Has VMS, since introduction of Alpha, ever
been capable of surpassing IBM mainframes for total throughput (not just CPU
cycles, IO as well). ? If not, will it be able with Galaxies ?
Or is there still anything that an IBM mainframe can do that a VMS box could
not ?
Arne Vajhøj
> If VMS is so robust and scalable and UNIX is unreliable, I am puzzled as to
> why one of the largest and most critical systems, the airline reservation
> system SABRE is switching to UNIX instead of going to VMS.
> If Galaxies is such a great advance and is unmatched in providing exactly what
> SABRE needs for its mission critical systems, why would it have lost such a
> sale ?
According to other posts, then Sabre is not switching from
IBM mainframe !
If they were, then VMS Galaxies was a good alternative.
> And ppppplease, stop using the Crédit Lyonnais as an example. It is old. Is
> that really the only one that they can use ?
The case does not get less good, because it is not brandnew.
Most stores like this never come out, because either they are
too small, or the company wants to keep a low profile in such
matters.
I can give a small example: a VMS system with shadowed disks, where
1) both disk towers were on the same UPS (not smart, but ...)
2) that UPS failed and stopped supplying power (not what you
expect from an UPS, but ...)
when I got power back on the system, then it continued fine,
no loss of data (transactions to a RDB database !). VMS
mount verification handled it just fine. How many of those
cheap systems can handle a power-fail on all disks including
system disk and come back again without reboot ?
Arne
Joshua Cope
> when I got power back on the system, then it continued fine,
> no loss of data (transactions to a RDB database !). VMS
> mount verification handled it just fine...
Mount Verification is your friend.
I've accidentally pulled the system disk out of a running system
before. No problem - pop it back in the box; a few seconds
of mount verification and everything continues as if nothing
happened...
------------------------------------------------------------
The above opinions and information are not necessarily
those of Compaq Computer Corporation.
------------------------------------------------------------
Mark Levy
could
>not ?
Yes. Run software written for the mainframes (without a lot of expensive
conversion, that is). Can you say JCL?
Mark Levy
System Management Associates, Inc.
JF Mezei
>
> >Or is there still anything that an IBM mainframe can do that a VMS box
> could
> >not ?
>
> Yes. Run software written for the mainframes (without a lot of expensive
> conversion, that is). Can you say JCL?
Actually, JCL is fairly straightforwardly converted into DCL.
DDNAMES = LOGICAL NAMES.
EXEC IDCAMS = CONVERT/FDL
etc etc etc.
It is when you move to a system which does not have the concept of a logical
name that your applications need to start to be changed.
zes...@my-dejanews.com
I still remember when one user was turning his body around in the computer
room and hit the write protect button of a RM03 drive (the system disk, of
course ;-).
No big deal - the system did give out a clear message on the operator
console and went into MV. After the write protect was removed, the system
gave another clear message and just went on. VMS (I think that was a late
V2 or early V3 release, so MV is nothing new!) saved another day ...
Many thanks for giving us mount verification!
--
Uwe Zessin
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
Phillip C. Thayer
bad power supply. It kept going off-line and back again all night (about 5 or
6 times) and when I came in the next morning and called the technician they
switched the power supply connection to another power supply. The disk
completed the MV, I shutdown the system normally, power supply was replaced
and system booted normally. I can tell you I was sweating pretty heavy but was
surprised at how resillient the MV really was.
Phillip
zes...@my-dejanews.com wrote:
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
"Committee - a group of men who individually can do
nothing but as a group decide that nothing can be done."
- Fred Allen
"A committee is a cul-de-sac down which ideas are
lured and then quietly strangled."
- Barnett Cocks, attributed
"A committee is an animal with four back legs."
- JOHN LE CARRÉ, Tinker Tailor Soldier Spy
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Bob Koehler
>VMS (I think that was a late
>V2 or early V3 release, so MV is nothing new!) saved another day ...
>
Mount verification was not part of VMS 2.
----------------------------------------------------------------------
Bob Koehler | Computer Sciences Corporation
Hubble Space Telescope Payload | Federal Sector, Civil Group
Flight Software Team | please remove ".aspm" when replying
Larry Kilgallen
> JF Mezei wrote:
>> And ppppplease, stop using the Crédit Lyonnais as an example. It is old. Is
>> that really the only one that they can use ?
>
> The case does not get less good, because it is not brandnew.
In fact it gets _better_ with age, since it is not a case of "this
is what the very latest VMS can do" but rather, "VMS has been doing
this for years". That is a particularly telling statement when no
other OS can do it yet.
Larry Kilgallen
Larry Kilgallen
> I still remember when one user was turning his body around in the computer
> room and hit the write protect button of a RM03 drive (the system disk, of
> course ;-).
>
> No big deal - the system did give out a clear message on the operator
> console and went into MV. After the write protect was removed, the system
> gave another clear message and just went on. VMS (I think that was a late
> V2 or early V3 release, so MV is nothing new!) saved another day ...
>
> Many thanks for giving us mount verification!
Andy Goldstein told the story at a DECUS LUG meeting some years back
that VMS Developers had thought about doing Mount Verification well
before they got around to implementing it. When one of their VMS
programmers took the system disk off line by mistake, that programmer
got the assignment to write mount verification !
Larry Kilgallen
Richard D. Piccard
I don't know how many times he told that story, so I may have heard it
at a different DECUS, but the way I recall it was that they were at that
time about halfway through the VMS re-build for the next test version,
at a time when the fastest VAX took a couple of days to build VMS, and he
took out the unit number plug (remember those). So the whole group of
developers was dead in the water.
Andy, do we remember this correctly?
RDP
--
==================================================================
Dick Piccard Academic Technology Manager
pic...@ohiou.edu Computer Services
http://oak.cats.ohiou.edu/~piccard/ Ohio University
Jerry Leichter
| >V2 or early V3 release, so MV is nothing new!) saved another day ...
| >
|
Agreed. In fact, I well recall the scenario described in an earlier
note (user bumps into the write-protect button on an RM03). This was a
VMS V2.4 (I think - close to that, anyway) system, and there was no
mount verification. A quick crash resulted.
(Conversely, you haven't lived 'till you've seen VMS power recovery. We
had a quick power glitch on the same 11/780. CPU dies; power to memory
stays. System re-boots. Console terminal - and every terminal on the
system - re-types the last prompt and most characters typed in response
- especially nifty to see and hear on the old LA hard-copy terminals.
Everything continues pretty much from where it left off. You never get
to see that these days, since systems tend to have full power backup, so
they don't reboot at all, or no separate power for the memory, so they
reboot cold.)
For those who don't remember the RM03, it had a hardware design flaw:
There were a series of indicated lights and buttons in an inset area at
the top front, but the inset wasn't quite deep enough, and the spring
and travel distance of the switches were very small - you could easily
trigger them by just brushing against the drive. I complained to our
hardware people, and learned that they had a whole project to come up
with a little plastic box with a spring-loaded cover that would fit into
the inset and protect the buttons. They had a project plan, engineering
drawings ... and a schedule that would have the first ones available in
about 6 months. So ... I took an old box, cut out a piece of cardboard,
cut a window in the middle (so you could see the indicator lights, and,
with a bit of duct tape, made a quick cover.
Everyone liked the "Leichter ECO" so much that it was applied to all
RM03's as they came through the door. The idea spread to other labs at
DEC from there. (The plastic box never got built.)
-- Jerry
Geoff Roberts
zes...@my-dejanews.com wrote in message <6vcg66$lv9$1...@nnrp1.dejanews.com>...
>In article <3618DE0C...@star.enet.dec.com>,
> co...@star.enet.dec.com wrote:
>> Arne Vajhøj wrote:
>>
>> > when I got power back on the system, then it continued fine,
>> > no loss of data (transactions to a RDB database !). VMS
>> > mount verification handled it just fine...
>>
>> Mount Verification is your friend.
AMEN!
I had a situation once where we had a momentary power failure, the UPS
inside the Vax 6310 kept the system running, but the
HSC50 that hosted ALL the drives rebooted immediately. There was an FTP
session open downloading something from somewhere at
the time. Mount verify brought all the drives back up when the HSC came
back on line (this takes a good ten minutes, it boots off tape)
then there was a flurry of disk activity, and all returned to normal. Even
more impressive was the fact that it kept the ftp session alive, buffering
the data as it came in, and writing it to the drive when it "reappeared", so
the ftp session, which finished downloading while the drives were still
offline, completed normally and didn't need to be redone.
Anyone want to try something like that on an NT box?? No?? Wonder why?
Just my 2c worth
Cheers
Geoff Roberts
Computer Systems Manager
Saint Mark's College
Port Pirie, South Australia
geof...@stmarksx.ppx.catholicx.edux.aux
Spam countermeasures,
remove the x's from email address before using
zes...@my-dejanews.com
Jerry Leichter <leic...@smarts.com> wrote:
> | >VMS (I think that was a late
> | >V2 or early V3 release, so MV is nothing new!) saved another day ...
> |
> | Mount verification was not part of VMS 2.
Thanks for the info, Bob. I wasn't a "VMS system manager" at that time -
and it's been a loooong time since then (3 major VMS releases ;-).
My primary task was PCB design - but I did some simple operator jobs, too.
I _did_ know what write-protect means and no - it wasn't me who hit
the button...
> Agreed. In fact, I well recall the scenario described in an earlier
> note (user bumps into the write-protect button on an RM03). This was a
> VMS V2.4 (I think - close to that, anyway) system, and there was no
> mount verification. A quick crash resulted.
When I acquired a new job some years ago one of my first actions was to
remove the '/NOMOUNT_VERIFICATION' qualifiers from the MOUNT commands!
**
The person who was put in charge didn't knew what MV was and wanted to
'go safe' ;-(
Ah yes, power failures and VMS. Same environment I was talking about, only
a few years later, when we had a VAX-8600 which was equiped with embedded
battery backup at our site:
When the power came back and the HSC50s had booted the machine happily
continued, except:
we were doing CAD (printed circuit board design) and the graphics was
done by some boxes connected to the UNIBUS. Too bad these 'boxes' didn't
have battery backup and needed a separate initialization program, so the
users lost their work despite battery battery on the VAX :-(
Unfortunately I see a similar trend today:
people put their VMS systems on UPS, but usually forget their network gear
which runs unprotected !!
Arne Vajhøj
> I can remember once when the system disk on a system I was working on had a
> bad power supply. It kept going off-line and back again all night (about 5 or
> 6 times) and when I came in the next morning and called the technician they
> switched the power supply connection to another power supply. The disk
> completed the MV, I shutdown the system normally, power supply was replaced
> and system booted normally. I can tell you I was sweating pretty heavy but was
> surprised at how resillient the MV really was.
The point is that those small daily stories does not hit the
press. They probably not even hit the boss's desk !
Missing problems are rarely noticed ! :-)
And that makes VMS relative anonymous among decision
makers. They do not hear about theese daily things, that
makes systems easier to operate.
Try to tell a CEO about the nice features of VMS BACKUP
copmpared with many other solutions. He will not get it, when
you start telling about CRC's and XOR groups etc..
Arne
Gary Mc Closkey
>Arne Vajhøj wrote:
>
>> when I got power back on the system, then it continued fine,
>> no loss of data (transactions to a RDB database !). VMS
>> mount verification handled it just fine...
>
>Mount Verification is your friend.
>
>before. No problem - pop it back in the box; a few seconds
>of mount verification and everything continues as if nothing
>happened...
No big deal - I did it on my amiga 1200 all the time. Even better, take out
one HD, and plug in another - although they had to have matching OS
versions, and system paths.
Gary.
--
gmc_...@indigo.ie. Remove underscore_duff to mail me.
Derek Fage
It would be intriguing if you tried to page in something from a =
pagefile
that on no way reflects the contents of what was paged out !
Derek...
> -----Original Message-----
> From: Gary Mc Closkey=20
> Sent: 07 October 1998 10:10
> To: Info...@Mvb.Saic.Com
> Subject: Re: VMS for those realy big critical applications
>=20
> .dec.com>...
> >Arne Vajh=F8j wrote:
> >
> >> when I got power back on the system, then it continued fine,
> >> no loss of data (transactions to a RDB database !). VMS
> >> mount verification handled it just fine...
> >
> >Mount Verification is your friend.
> >
> >I've accidentally pulled the system disk out of a running system
> >before. No problem - pop it back in the box; a few seconds
> >of mount verification and everything continues as if nothing
> >happened...
>=20
> No big deal - I did it on my amiga 1200 all the time. Even better,
> take out
> one HD, and plug in another - although they had to have matching OS
> versions, and system paths.
>=20
> Gary.
>=20
>=20
> --
> gmc_...@indigo.ie. Remove underscore_duff to mail me.
>=20
>=20
Bob Koehler
In article <361A76...@smarts.com>, Jerry Leichter <leic...@smarts.com> writes:
>
>For those who don't remember the RM03,
> So ... I took an old box, cut out a piece of cardboard,
>cut a window in the middle (so you could see the indicator lights, and,
>with a bit of duct tape, made a quick cover.
>
>Everyone liked the "Leichter ECO" so much that it was applied to all
>RM03's as they came through the door. The idea spread to other labs at
>DEC from there. (The plastic box never got built.)
>
RM03 wan't the last. We had the same problem with RF series removables.
A system manager we knew found a company making plexiglass covers held
on with velcro, sutom fit to each model for about $12 (US) each. We just
went down to the store and grabbed some plexiglass and velcro, faster nd
cheaper.
Mark D. Jilson
units with covers over the switches. I don't remember having many problems
w/ the switches and we'd have close to a hundred units on the floor under
going final burn-in testing at any one time.
Jerry Leichter wrote:
> For those who don't remember the RM03, it had a hardware design flaw:
> There were a series of indicated lights and buttons in an inset area at
> the top front, but the inset wasn't quite deep enough, and the spring
> and travel distance of the switches were very small - you could easily
> trigger them by just brushing against the drive. I complained to our
> hardware people, and learned that they had a whole project to come up
> with a little plastic box with a spring-loaded cover that would fit into
> the inset and protect the buttons. They had a project plan, engineering
> drawings ... and a schedule that would have the first ones available in
> about 6 months. So ... I took an old box, cut out a piece of cardboard,
> cut a window in the middle (so you could see the indicator lights, and,
> with a bit of duct tape, made a quick cover.
>
> Everyone liked the "Leichter ECO" so much that it was applied to all
> RM03's as they came through the door. The idea spread to other labs at
> DEC from there. (The plastic box never got built.)
--
Jilly - Working from Home in the Chemung River Valley - Lockwood, NY
- ji...@clarityconnect.com - Brett Bodine fan
- m_ji...@csc32.enet.dec.com - since 1975 or so
- http://www.clarityconnect.com/webpages2/jilly -
Alan Greig
Jerry Leichter wrote:
> (Conversely, you haven't lived 'till you've seen VMS power recovery. We
> had a quick power glitch on the same 11/780. CPU dies; power to memory
> stays. System re-boots. Console terminal - and every terminal on the
> system - re-types the last prompt and most characters typed in response
> - especially nifty to see and hear on the old LA hard-copy terminals.
> Everything continues pretty much from where it left off. You never get
> to see that these days, since systems tend to have full power backup, so
> they don't reboot at all, or no separate power for the memory, so they
> reboot cold.)
>
DEC were always good at that. I remember roughly the following sequence on
a DEC-20 circa 1980 although I can no longer recall the exact messages.
XMAS eve OPS go off for a break until after New Year (long Scottish break)
and two hours later the power fails and circuit breakers trip. Ten days
later OPS return and reset machine room breakers and console prints
something like this:
%DECSYSTEM-20 NOT RUNNING
POWER FAIL RESTART
ATTEMPTING TO CONTINUE SYSTEM... SUCCESS
BUGCHK PWRRES
ATTEMPTING POWER FAIL RESTART
%DECSYSTEM-20 NOT RUNNING
[SY0: REDIRECTED TO PD0:]
[DECSystem-20 Continued]
All disks spin up,batch jobs continue, and terminal sessions on local lines
resume (because we'd patched the TTDALL routine called during the power fail
restart due to a spate of transient power fails). Never expected a transient
power fail spanning almost two weeks though! Not long after the old core
memory was upgraded to MOS memory which suffers amnesia without the power
and things were never quite the same.
--
Alan Greig
Alan Greig
Larry Kilgallen wrote:
Come to think of it TOPS-20 would even allow you to remove a pack from a failed drive
and load it into another one. Assuming you were sure it was the drive and not the pack
at fault of course... So long ago.
>
> Larry Kilgallen
--
Alan Greig
Phillip C. Thayer
neck in the computer room. It was a relatively small company (much larger now) that
is an EDI VAN. The CEO was so impressed that he went right out and ordered a newer
VAX (before Alphas) to put in. He was an NCR diehard before that. Now that company
is running on exclusively Alpha hardware.
Phillip
Arne Vajhøj wrote:
--
Wayne W. Scott
Can you say what company it is and what they do?
Thanks,
Wayne
=====