| The ICL 2900 [message #336191] |
Sun, 22 January 2017 18:49  |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
I had been interested in finding out about the ICL 1900, because it was a
computer with a 24-bit word.
There are some manuals for it on Bitsavers now, but where I first found
information about it was on a different site.
I happened to glance at a Bible, in the Revised Standard Version, which had small illustrations in with the text, and which was also unusual in that some parts were in smaller type and in three columns, while most of the text was in two.
At the start was a note that the book had been typeset by an I.C.T. 1500 computer.
I looked it up, and found that it was a copy of the RCA 301, made under license.
But that led me to look at other items about the history of ICT and ICL. And then I learned that the ICL 2900 was a computer designed as a successor to the series of machines which was based on the RCA Spectra computers.
A computer designed to be better than the IBM 360, due to using more advanced concepts of design, sounded interesting.
So I did a web search, and found two documents on Fujitsu's web site in PDF
format.
One was a book, "The ICL 2900 Series", by J. K. Buckle.
Another was an issue of the ICL Technical Journal - Volume 7, Number 2,
November 1990.
The latter was typeset conventionally - but I noticed that the former, the buck
by J. K. Buckle, was done on some form of the IBM Selectric Composer.
So we come full circle as far as typesetting is concerned.
In any event, the ICL Technical Journal describes an SX series of machines,
which was claimed to be more efficient than an Amdahl 5800 series machine
despite being made with the same hardware technology from Fujitsu.
As to the ICL 2900 architecture, it was a lot like the Burroughs machines, but
it used IBM 360-like datatypes, not a 48-bit word like they did, and instead of
being focused on one language (ALGOL 60), it was designed to work equally well
with any higher-level language.
John Savard
|
|
|
|
| Re: The ICL 2900 [message #336193 is a reply to message #336191] |
Sun, 22 January 2017 19:24  |
|
|
Originally posted by: Bob Eager
On Sun, 22 Jan 2017 15:49:33 -0800, Quadibloc wrote:
> I had been interested in finding out about the ICL 1900, because it was
> a computer with a 24-bit word.
>
> There are some manuals for it on Bitsavers now, but where I first found
> information about it was on a different site.
>
> I happened to glance at a Bible, in the Revised Standard Version, which
> had small illustrations in with the text, and which was also unusual in
> that some parts were in smaller type and in three columns, while most of
> the text was in two.
>
> At the start was a note that the book had been typeset by an I.C.T. 1500
> computer.
>
> I looked it up, and found that it was a copy of the RCA 301, made under
> license.
>
> But that led me to look at other items about the history of ICT and ICL.
> And then I learned that the ICL 2900 was a computer designed as a
> successor to the series of machines which was based on the RCA Spectra
> computers.
>
> A computer designed to be better than the IBM 360, due to using more
> advanced concepts of design, sounded interesting.
>
> So I did a web search, and found two documents on Fujitsu's web site in
> PDF format.
>
> One was a book, "The ICL 2900 Series", by J. K. Buckle.
>
> Another was an issue of the ICL Technical Journal - Volume 7, Number 2,
> November 1990.
>
> The latter was typeset conventionally - but I noticed that the former,
> the buck by J. K. Buckle, was done on some form of the IBM Selectric
> Composer.
>
> So we come full circle as far as typesetting is concerned.
>
> In any event, the ICL Technical Journal describes an SX series of
> machines,
> which was claimed to be more efficient than an Amdahl 5800 series
> machine despite being made with the same hardware technology from
> Fujitsu.
>
> As to the ICL 2900 architecture, it was a lot like the Burroughs
> machines, but it used IBM 360-like datatypes, not a 48-bit word like
> they did, and instead of being focused on one language (ALGOL 60), it
> was designed to work equally well with any higher-level language.
Loosely based on the Manchester MU5. It had a single accumulator and a
single index register, was stack and descriptor based, and had caching on
the 'business' end of the stack to compensate for the lack of registers.
An easy machine to write compilers for, as I know!
The accumulator was variable size - 32, 64 and 128 bits. One could choose
the size when it had something loaded into it. Certain multiplication and
division operations would alter the size, as might be expected.
Full 32 bit addressing, segmented and paged. 1kB pages and 256kB
segments. Half the segments were 'public' (shared between processes) and
half were 'private' (process-local).
Ring based security with 16 Access Control Levels via a special register
(ACR).
It could also assume two other 'personalities', and switch betwqeen them.
One was the ICL 1900 (the previous 24 bit system). The other was the
English Electric System 4, so not dissimilar to a 360.
If anyone wants to know more, please ask. I worked on one for several
years, and helped in the development of a home-brew operating system. I
also ended up patching the microcode to get round a hardware design
error. For more details on that, see:
http://www.bobeager.uk/anecdotes.html
and scroll down a bit.
I am currently working on a 2900 emulator.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| ICL 1900 and 2900 computers. [message #336194 is a reply to message #336191] |
Sun, 22 January 2017 20:22 |
Anonymous
Messages: 58
Registered: July 2012
Karma: 0
|
Member |
|
|
Quadibloc wrote:
> I had been interested in finding out about the ICL 1900, because it was a
> computer with a 24-bit word.
See:
http://www.ourcomputerheritage.org/ict_15.htm
http://www.icl1900.co.uk
and
http://www.findlayw.plus.com/ICL1900/index.html
The last of these contains the second (and vastly better, not being based
on a Zurich original)
ICL 1900 Pascal compiler.
BTW, Bob, Delwyn Holroyd at TNMoC, Bletchley Park, has a MT I donated that
contains 2900 Pascal.
There are some problems reading the object files due to read errors that
Delwyn thinks may be correctable,
but I have managed to reconstruct seemingly valid source texts of the
compiler, diagnostic system,
and related SCL macros.
|
|
|
|
| Re: ICL 1900 and 2900 computers. [message #336195 is a reply to message #336194] |
Sun, 22 January 2017 20:25 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 01:22:33 +0000, Anonymous wrote:
> Quadibloc wrote:
>
>> I had been interested in finding out about the ICL 1900, because it was
>> a computer with a 24-bit word.
>
> See:
>
> http://www.ourcomputerheritage.org/ict_15.htm
>
> http://www.icl1900.co.uk
>
> and
>
> http://www.findlayw.plus.com/ICL1900/index.html
>
> The last of these contains the second (and vastly better, not being
> based on a Zurich original)
> ICL 1900 Pascal compiler.
>
> BTW, Bob, Delwyn Holroyd at TNMoC, Bletchley Park, has a MT I donated
> that contains 2900 Pascal.
> There are some problems reading the object files due to read errors that
> Delwyn thinks may be correctable,
> but I have managed to reconstruct seemingly valid source texts of the
> compiler, diagnostic system,
> and related SCL macros.
Thanks - I know Delwyn, and I'll ask him when I see him next (probably in
the pub!)
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
|
|
| Re: ICL 1900 and 2900 computers. [message #336197 is a reply to message #336195] |
Sun, 22 January 2017 20:29 |
Bill Findlay
Messages: 286
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Bob Eager <news0006@eager.cx> wrote:
> On Mon, 23 Jan 2017 01:22:33 +0000, Anonymous wrote:
>
>> Quadibloc wrote:
>>
>>> I had been interested in finding out about the ICL 1900, because it was
>>> a computer with a 24-bit word.
>>
>> See:
>>
>> http://www.ourcomputerheritage.org/ict_15.htm
>>
>> http://www.icl1900.co.uk
>>
>> and
>>
>> http://www.findlayw.plus.com/ICL1900/index.html
>>
>> The last of these contains the second (and vastly better, not being
>> based on a Zurich original)
>> ICL 1900 Pascal compiler.
>>
>> BTW, Bob, Delwyn Holroyd at TNMoC, Bletchley Park, has a MT I donated
>> that contains 2900 Pascal.
>> There are some problems reading the object files due to read errors that
>> Delwyn thinks may be correctable,
>> but I have managed to reconstruct seemingly valid source texts of the
>> compiler, diagnostic system, and related SCL macros.
>
> Thanks - I know Delwyn, and I'll ask him when I see him next (probably in the pub!)
If you want the 2900 sources, let me know.
--
Bill Findlay
|
|
|
|
| Re: The ICL 2900 [message #336213 is a reply to message #336191] |
Mon, 23 January 2017 09:54 |
Stephen Wolstenholme
Messages: 231
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
worked on the ICL 2900 and then it's replacement S39. On the customer
sites I was working on the ICL systems were faster than IBM systems
doing the same jobs but the IBM machines were more reliable. An
interesting thing I noticed was that most customers went for neither
ICL or IBM when the systems were replaced.
Steve
--
Neural Network Software for Windows http://www.npsnn.com
|
|
|
|
| Re: The ICL 2900 [message #336215 is a reply to message #336193] |
Mon, 23 January 2017 11:22 |
|
|
Originally posted by: albiorix
On Monday, January 23, 2017 at 2:24:41 AM UTC+2, Bob Eager wrote:
> On Sun, 22 Jan 2017 15:49:33 -0800, Quadibloc wrote:
>
>> I had been interested in finding out about the ICL 1900, because it was
>> a computer with a 24-bit word.
>>
>> There are some manuals for it on Bitsavers now, but where I first found
>> information about it was on a different site.
>>
>> I happened to glance at a Bible, in the Revised Standard Version, which
>> had small illustrations in with the text, and which was also unusual in
>> that some parts were in smaller type and in three columns, while most of
>> the text was in two.
>>
>> At the start was a note that the book had been typeset by an I.C.T. 1500
>> computer.
>>
>> I looked it up, and found that it was a copy of the RCA 301, made under
>> license.
>>
>> But that led me to look at other items about the history of ICT and ICL..
>> And then I learned that the ICL 2900 was a computer designed as a
>> successor to the series of machines which was based on the RCA Spectra
>> computers.
>>
>> A computer designed to be better than the IBM 360, due to using more
>> advanced concepts of design, sounded interesting.
>>
>> So I did a web search, and found two documents on Fujitsu's web site in
>> PDF format.
>>
>> One was a book, "The ICL 2900 Series", by J. K. Buckle.
>>
>> Another was an issue of the ICL Technical Journal - Volume 7, Number 2,
>> November 1990.
>>
>> The latter was typeset conventionally - but I noticed that the former,
>> the buck by J. K. Buckle, was done on some form of the IBM Selectric
>> Composer.
>>
>> So we come full circle as far as typesetting is concerned.
>>
>> In any event, the ICL Technical Journal describes an SX series of
>> machines,
>> which was claimed to be more efficient than an Amdahl 5800 series
>> machine despite being made with the same hardware technology from
>> Fujitsu.
>>
>> As to the ICL 2900 architecture, it was a lot like the Burroughs
>> machines, but it used IBM 360-like datatypes, not a 48-bit word like
>> they did, and instead of being focused on one language (ALGOL 60), it
>> was designed to work equally well with any higher-level language.
>
> Loosely based on the Manchester MU5. It had a single accumulator and a
> single index register, was stack and descriptor based, and had caching on
> the 'business' end of the stack to compensate for the lack of registers.
> An easy machine to write compilers for, as I know!
>
> The accumulator was variable size - 32, 64 and 128 bits. One could choose
> the size when it had something loaded into it. Certain multiplication and
> division operations would alter the size, as might be expected.
>
> Full 32 bit addressing, segmented and paged. 1kB pages and 256kB
> segments. Half the segments were 'public' (shared between processes) and
> half were 'private' (process-local).
>
> Ring based security with 16 Access Control Levels via a special register
> (ACR).
>
> It could also assume two other 'personalities', and switch betwqeen them.
> One was the ICL 1900 (the previous 24 bit system). The other was the
> English Electric System 4, so not dissimilar to a 360.
>
> If anyone wants to know more, please ask. I worked on one for several
> years, and helped in the development of a home-brew operating system. I
> also ended up patching the microcode to get round a hardware design
> error. For more details on that, see:
>
> http://www.bobeager.uk/anecdotes.html
>
> and scroll down a bit.
>
> I am currently working on a 2900 emulator.
>
>
> --
> Using UNIX since v6 (1975)...
>
> Use the BIG mirror service in the UK:
> http://www.mirrorservice.org
I worked on the 2900 and later, the 3900 in the 80s, - initially at a large chain store, but later as part of a team which ported the Software-AG products (ADABAS and Natural) from the IBM-370 type machines to the ICL kit. As a result we were one of the few customers to work with S3 and SFL (ICLs "assembly language").
That was a load of fun. I'd enjoy playing with that again if an emulator ever happened.
Of course, getting the OS and software might be problematic.
|
|
|
|
| Re: The ICL 2900 [message #336217 is a reply to message #336215] |
Mon, 23 January 2017 12:55 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 08:22:59 -0800, albiorix wrote:
> I worked on the 2900 and later, the 3900 in the 80s, - initially at a
> large chain store, but later as part of a team which ported the
> Software-AG products (ADABAS and Natural) from the IBM-370 type machines
> to the ICL kit. As a result we were one of the few customers to work
> with S3 and SFL (ICLs "assembly language").
I used SFL quite a lot. I even have a manual somewhere.
> That was a load of fun. I'd enjoy playing with that again if an
> emulator ever happened.
>
> Of course, getting the OS and software might be problematic.
I would be using the homebrew one.
If you look at the website, you'll see we were doomed to use VME/K.
Rolling 13 week MTBF of the system was about 20 hours. With our own
system, it was 2000 hours.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336218 is a reply to message #336213] |
Mon, 23 January 2017 12:56 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 14:54:44 +0000, Stephen Wolstenholme wrote:
> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
> worked on the ICL 2900 and then it's replacement S39. On the customer
> sites I was working on the ICL systems were faster than IBM systems
> doing the same jobs but the IBM machines were more reliable. An
> interesting thing I noticed was that most customers went for neither ICL
> or IBM when the systems were replaced.
Reliability (hardware wise) was dreadful. A lot of the stuff in our
operating system was recovery from transient errors.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336222 is a reply to message #336213] |
Mon, 23 January 2017 15:40 |
Peter Flass
Messages: 8375
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
Stephen Wolstenholme <steve@easynn.com> wrote:
> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
> worked on the ICL 2900 and then it's replacement S39. On the customer
> sites I was working on the ICL systems were faster than IBM systems
> doing the same jobs but the IBM machines were more reliable. An
> interesting thing I noticed was that most customers went for neither
> ICL or IBM when the systems were replaced.
Where did they go?
--
Pete
|
|
|
|
| Re: The ICL 2900 [message #336227 is a reply to message #336222] |
Mon, 23 January 2017 16:18 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 13:40:03 -0700, Peter Flass wrote:
> Stephen Wolstenholme <steve@easynn.com> wrote:
>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>> worked on the ICL 2900 and then it's replacement S39. On the customer
>> sites I was working on the ICL systems were faster than IBM systems
>> doing the same jobs but the IBM machines were more reliable. An
>> interesting thing I noticed was that most customers went for neither
>> ICL or IBM when the systems were replaced.
>
> Where did they go?
We went for DEC; they won the tender.
IBM also bid but had an incompetent sales team.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336228 is a reply to message #336191] |
Mon, 23 January 2017 17:11 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 22:05:05 +0000, Huge wrote:
> On 2017-01-23, Bob Eager <news0006@eager.cx> wrote:
>> On Mon, 23 Jan 2017 13:40:03 -0700, Peter Flass wrote:
>>
>>> Stephen Wolstenholme <steve@easynn.com> wrote:
>>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>>> sites I was working on the ICL systems were faster than IBM systems
>>>> doing the same jobs but the IBM machines were more reliable. An
>>>> interesting thing I noticed was that most customers went for neither
>>>> ICL or IBM when the systems were replaced.
>>>
>>> Where did they go?
>>
>> We went for DEC; they won the tender.
>>
>> IBM also bid but had an incompetent sales team.
>
> There's a sentence I never thought I'd see; an incompetent IBM sales
> team!
I think it was their first account. It was wide open and should have been
easy for them.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336229 is a reply to message #336213] |
Mon, 23 January 2017 17:17 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme wrote:
> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
> worked on the ICL 2900 and then it's replacement S39. On the customer
> sites I was working on the ICL systems were faster than IBM systems
> doing the same jobs but the IBM machines were more reliable. An
> interesting thing I noticed was that most customers went for neither
> ICL or IBM when the systems were replaced.
Back then some customers had emotional prejudices which influenced
their selection of computer. As we know, many customers felt, "you
never get fired for buying IBM", and went that way, even if the
price/performance/quality might have been better with alternatives.
On the other hand, there were some customers who disliked IBM for
various reasons, and refused to consider an IBM machine.
|
|
|
|
| Re: The ICL 2900 [message #336232 is a reply to message #336229] |
Mon, 23 January 2017 18:09 |
dave[1][2]
Messages: 119
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
On 23/01/2017 22:17, hancock4@bbs.cpcn.com wrote:
> On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme wrote:
>
>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>> worked on the ICL 2900 and then it's replacement S39. On the customer
>> sites I was working on the ICL systems were faster than IBM systems
>> doing the same jobs but the IBM machines were more reliable. An
>> interesting thing I noticed was that most customers went for neither
>> ICL or IBM when the systems were replaced.
>
> Back then some customers had emotional prejudices which influenced
> their selection of computer. As we know, many customers felt, "you
> never get fired for buying IBM", and went that way, even if the
> price/performance/quality might have been better with alternatives.
> On the other hand, there were some customers who disliked IBM for
> various reasons, and refused to consider an IBM machine.
>
In general you could get better price/performance elsewhere. IBM's
technique was to sell to high management, and convince them. The
technical sales teams were not used to having to work.
I remember when the company I worked for was looking at buying a new
machine around the same time as ICL2900 appeared, IBM suggested that we
run on-line during the day and batch overnight. Went down like a lead
balloon with management.
We were Honeywell and bought a Honeywell L66/10 and did run batch and
on-line at the same time, albeit after a memory upgrade...
Dave
G4UGM
|
|
|
|
| Re: The ICL 2900 [message #336233 is a reply to message #336232] |
Mon, 23 January 2017 18:31 |
|
|
Originally posted by: Bob Eager
On Mon, 23 Jan 2017 23:09:41 +0000, David Wade wrote:
> On 23/01/2017 22:17, hancock4@bbs.cpcn.com wrote:
>> On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme
>> wrote:
>>
>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>> sites I was working on the ICL systems were faster than IBM systems
>>> doing the same jobs but the IBM machines were more reliable. An
>>> interesting thing I noticed was that most customers went for neither
>>> ICL or IBM when the systems were replaced.
>>
>> Back then some customers had emotional prejudices which influenced
>> their selection of computer. As we know, many customers felt, "you
>> never get fired for buying IBM", and went that way, even if the
>> price/performance/quality might have been better with alternatives.
>> On the other hand, there were some customers who disliked IBM for
>> various reasons, and refused to consider an IBM machine.
>>
>>
> In general you could get better price/performance elsewhere. IBM's
> technique was to sell to high management, and convince them. The
> technical sales teams were not used to having to work.
>
> I remember when the company I worked for was looking at buying a new
> machine around the same time as ICL2900 appeared, IBM suggested that we
> run on-line during the day and batch overnight. Went down like a lead
> balloon with management.
Our 2900 was really very underpowered. It could just about run one batch
stream and struggled to support 32 on-line users at the same time. There
were some terrible short cuts in the operating system (e.g. "If in doubt,
crash")
When we changed to the homebrew system, the MTBF increased 100-fold. It
also supported two batch streams and 48 on-line users.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336243 is a reply to message #336222] |
Tue, 24 January 2017 04:47 |
Stephen Wolstenholme
Messages: 231
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Mon, 23 Jan 2017 13:40:03 -0700, Peter Flass
<peter_flass@yahoo.com> wrote:
> Stephen Wolstenholme <steve@easynn.com> wrote:
>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>> worked on the ICL 2900 and then it's replacement S39. On the customer
>> sites I was working on the ICL systems were faster than IBM systems
>> doing the same jobs but the IBM machines were more reliable. An
>> interesting thing I noticed was that most customers went for neither
>> ICL or IBM when the systems were replaced.
>
> Where did they go?
It varied. Of the ones I stayed in contact with, one went with
Honeywell but is now back with IBM, one went with Univac and then
Fujitsu. The customer site where I worked the longest is now a car
park!
Steve
--
Neural Network Software for Windows http://www.npsnn.com
|
|
|
|
| Re: The ICL 2900 [message #336247 is a reply to message #336229] |
Tue, 24 January 2017 09:02 |
Peter Flass
Messages: 8375
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
<hancock4@bbs.cpcn.com> wrote:
> On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme wrote:
>
>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>> worked on the ICL 2900 and then it's replacement S39. On the customer
>> sites I was working on the ICL systems were faster than IBM systems
>> doing the same jobs but the IBM machines were more reliable. An
>> interesting thing I noticed was that most customers went for neither
>> ICL or IBM when the systems were replaced.
>
> Back then some customers had emotional prejudices which influenced
> their selection of computer. As we know, many customers felt, "you
> never get fired for buying IBM", and went that way, even if the
> price/performance/quality might have been better with alternatives.
> On the other hand, there were some customers who disliked IBM for
> various reasons, and refused to consider an IBM machine.
>
IMO DEC was getting into the same league just before the end. The equipment
was about as reliable, the software was good and covered a great number of
areas customers wanted. Field service was excellent. Sales were OK, but it
was still much of "do it yourself". That's what makes Dec's demise that
much sadder.
--
Pete
|
|
|
|
| Re: The ICL 2900 [message #336248 is a reply to message #336247] |
Tue, 24 January 2017 09:09 |
|
|
Originally posted by: Bob Eager
On Tue, 24 Jan 2017 07:02:35 -0700, Peter Flass wrote:
> <hancock4@bbs.cpcn.com> wrote:
>> On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme
>> wrote:
>>
>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>> sites I was working on the ICL systems were faster than IBM systems
>>> doing the same jobs but the IBM machines were more reliable. An
>>> interesting thing I noticed was that most customers went for neither
>>> ICL or IBM when the systems were replaced.
>>
>> Back then some customers had emotional prejudices which influenced
>> their selection of computer. As we know, many customers felt, "you
>> never get fired for buying IBM", and went that way, even if the
>> price/performance/quality might have been better with alternatives.
>> On the other hand, there were some customers who disliked IBM for
>> various reasons, and refused to consider an IBM machine.
>>
>>
> IMO DEC was getting into the same league just before the end. The
> equipment was about as reliable, the software was good and covered a
> great number of areas customers wanted. Field service was excellent.
> Sales were OK, but it was still much of "do it yourself". That's what
> makes Dec's demise that much sadder.
Yes, we got a VAXcluster (the base CPU power of an 8800, running SMP, was
a bit under the IBM offering, so they chucked in a couple of 8200s). In
fact, since it was on VMS 4.6, it couldn't do SMP anyway, and didn't for
years.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336249 is a reply to message #336247] |
Tue, 24 January 2017 09:43 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Peter Flass wrote:
> <hancock4@bbs.cpcn.com> wrote:
>> On Monday, January 23, 2017 at 9:55:13 AM UTC-5, Stephen Wolstenholme
wrote:
>>
>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>> sites I was working on the ICL systems were faster than IBM systems
>>> doing the same jobs but the IBM machines were more reliable. An
>>> interesting thing I noticed was that most customers went for neither
>>> ICL or IBM when the systems were replaced.
>>
>> Back then some customers had emotional prejudices which influenced
>> their selection of computer. As we know, many customers felt, "you
>> never get fired for buying IBM", and went that way, even if the
>> price/performance/quality might have been better with alternatives.
>> On the other hand, there were some customers who disliked IBM for
>> various reasons, and refused to consider an IBM machine.
>>
>
> IMO DEC was getting into the same league just before the end. The equipment
> was about as reliable, the software was good and covered a great number of
> areas customers wanted. Field service was excellent. Sales were OK, but it
> was still much of "do it yourself". That's what makes Dec's demise that
> much sadder.
>
DEC's customers wanted "do it yourself". That's why they bought DEC
instead of IBM. IBM didn't allow cusotmers to tweak anything. So
DEC sold the hardware, providing the monitor software and other
OS "tools" so that the customer could develope or run whatever
they wanted. They could also build or buy other kinds of hardware
and include the software support in the monitor and CUSPs because
we shipped the sources. Compuserv, ADP, ORNL and First Data all had
their own tweaks which were applicable to their business (but
not to other kinds of computing).
Then the -20 happened...
/BAH
|
|
|
|
| Re: The ICL 2900 [message #336250 is a reply to message #336218] |
Tue, 24 January 2017 09:43 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Bob Eager wrote:
> On Mon, 23 Jan 2017 14:54:44 +0000, Stephen Wolstenholme wrote:
>
>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>> worked on the ICL 2900 and then it's replacement S39. On the customer
>> sites I was working on the ICL systems were faster than IBM systems
>> doing the same jobs but the IBM machines were more reliable. An
>> interesting thing I noticed was that most customers went for neither ICL
>> or IBM when the systems were replaced.
>
> Reliability (hardware wise) was dreadful. A lot of the stuff in our
> operating system was recovery from transient errors.
Hardware was always dreadful back then. It was up to the software to
deal with it and make the system work well despite the hardware
limitations.
/BAH
|
|
|
|
| Re: The ICL 2900 [message #336251 is a reply to message #336228] |
Tue, 24 January 2017 09:43 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Bob Eager wrote:
> On Mon, 23 Jan 2017 22:05:05 +0000, Huge wrote:
>
>> On 2017-01-23, Bob Eager <news0006@eager.cx> wrote:
>>> On Mon, 23 Jan 2017 13:40:03 -0700, Peter Flass wrote:
>>>
>>>> Stephen Wolstenholme <steve@easynn.com> wrote:
>>>> > In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>>> > worked on the ICL 2900 and then it's replacement S39. On the customer
>>>> > sites I was working on the ICL systems were faster than IBM systems
>>>> > doing the same jobs but the IBM machines were more reliable. An
>>>> > interesting thing I noticed was that most customers went for neither
>>>> > ICL or IBM when the systems were replaced.
>>>>
>>>> Where did they go?
>>>
>>> We went for DEC; they won the tender.
>>>
>>> IBM also bid but had an incompetent sales team.
>>
>> There's a sentence I never thought I'd see; an incompetent IBM sales
>> team!
>
> I think it was their first account. It was wide open and should have been
> easy for them.
>
IBM sales people could be arrogant because they didn't believe they had
any competition.
/BAH
|
|
|
|
| Re: The ICL 2900 [message #336252 is a reply to message #336250] |
Tue, 24 January 2017 10:34 |
|
|
Originally posted by: Bob Eager
On Tue, 24 Jan 2017 14:43:01 +0000, jmfbahciv wrote:
> Bob Eager wrote:
>> On Mon, 23 Jan 2017 14:54:44 +0000, Stephen Wolstenholme wrote:
>>
>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>> sites I was working on the ICL systems were faster than IBM systems
>>> doing the same jobs but the IBM machines were more reliable. An
>>> interesting thing I noticed was that most customers went for neither
>>> ICL or IBM when the systems were replaced.
>>
>> Reliability (hardware wise) was dreadful. A lot of the stuff in our
>> operating system was recovery from transient errors.
>
> Hardware was always dreadful back then. It was up to the software to
> deal with it and make the system work well despite the hardware
> limitations.
Well, no. The reliabity of our DEC stuff was good. And the reliability of
that machine's immediate predecessor was a lot better too.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336253 is a reply to message #336251] |
Tue, 24 January 2017 10:35 |
|
|
Originally posted by: Bob Eager
On Tue, 24 Jan 2017 14:43:03 +0000, jmfbahciv wrote:
> Bob Eager wrote:
>> On Mon, 23 Jan 2017 22:05:05 +0000, Huge wrote:
>>
>>> On 2017-01-23, Bob Eager <news0006@eager.cx> wrote:
>>>> On Mon, 23 Jan 2017 13:40:03 -0700, Peter Flass wrote:
>>>>
>>>> > Stephen Wolstenholme <steve@easynn.com> wrote:
>>>> >> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>>> >> worked on the ICL 2900 and then it's replacement S39. On the
>>>> >> customer sites I was working on the ICL systems were faster than
>>>> >> IBM systems doing the same jobs but the IBM machines were more
>>>> >> reliable. An interesting thing I noticed was that most customers
>>>> >> went for neither ICL or IBM when the systems were replaced.
>>>> >
>>>> > Where did they go?
>>>>
>>>> We went for DEC; they won the tender.
>>>>
>>>> IBM also bid but had an incompetent sales team.
>>>
>>> There's a sentence I never thought I'd see; an incompetent IBM sales
>>> team!
>>
>> I think it was their first account. It was wide open and should have
>> been easy for them.
>>
> IBM sales people could be arrogant because they didn't believe they had
> any competition.
These two (women) were not arrogant, just inexperienced.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336254 is a reply to message #336249] |
Tue, 24 January 2017 11:06 |
scott
Messages: 4237
Registered: February 2012
Karma: 0
|
Senior Member |
|
|
jmfbahciv <See.above@aol.com> writes:
> Peter Flass wrote:
>>
>> IMO DEC was getting into the same league just before the end. The equipment
>> was about as reliable, the software was good and covered a great number of
>> areas customers wanted. Field service was excellent. Sales were OK, but it
>> was still much of "do it yourself". That's what makes Dec's demise that
>> much sadder.
>>
> DEC's customers wanted "do it yourself". That's why they bought DEC
> instead of IBM. IBM didn't allow cusotmers to tweak anything.
I'm sorry - but it was quite common for IBM customers to
tweek the OS. Significantly in many cases. I don't doubt that
many also had custom block-mux channels for various purposes.
|
|
|
|
| Re: The ICL 2900 [message #336255 is a reply to message #336248] |
Tue, 24 January 2017 12:44 |
Anne & Lynn Wheel
Messages: 3156
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Bob Eager <news0006@eager.cx> writes:
> Yes, we got a VAXcluster (the base CPU power of an 8800, running SMP, was
> a bit under the IBM offering, so they chucked in a couple of 8200s). In
> fact, since it was on VMS 4.6, it couldn't do SMP anyway, and didn't for
> years.
old reference to announce of "real" VMS SMP ("DEC Stalks Big Game with
Symmetrical VMS")
http://www.garlic.com/~lynn/2007.html#email880329
old reference with decade of vax sales, sliced&diced by year, model, etc
(8800, 30-1986, 170-1987) http://www.garlic.com/~lynn/2002f.html#0
We had started IBM HA/CMP project with RS/6000
http://www.garlic.com/~lynn/subtopic.html#hacmp
.... including working with some of the "open system" RDBMS vendors, old
reference to Jan1992 meeting in ellison's conference room on cluster
scaleup http://www.garlic.com/~lynn/95.html#13
that had VMS/cluster support in their same source base ... to ease port,
I did API for global lock manager that emulated the VMS/cluster
semantics. The implementation also benefited from list that the RDBMS
vendors had about what VMS/cluster could have done better.
Then at ACM SIGOPS conference, I got in dustup with Jim Gray ... a
decade earlier, I had worked with Jim at San Jose Research on the
original SQL/relational implementation
http://www.garlic.com/~lynn/submain.html#systemr
the dust-up was over whether you could do high-availability with
"commodity" systems (at the time he was working for DEC DBMS). Then when
DEC DBMS was sold to Oracle, he took a sabbatical ... and then shows up
as head of Microsoft San Francisco Research. I got my "revenge" when he
shows up on stage with CEO of Microsoft announcing microsoft
high-availability on intel platforms.
As I've mentioned before, within a few weeks of the Jan1992 Ellison
meeting, cluster scaleup was transferred, announced as IBM supercomputer
for technical/scientific *ONLY* (we had also been working with national
labs on cluster scaleup uses), and we were told we couldn't work on
anything with more than four processors. Contributing was that the
mainframe DB2 group were complaining if I was allowed to go ahead, it
would be at least 5yrs ahead of what they were doing.
--
virtualization experience starting Jan1968, online at home since Mar1970
|
|
|
|
| Re: The ICL 2900 [message #336257 is a reply to message #336254] |
Tue, 24 January 2017 13:17 |
Anne & Lynn Wheel
Messages: 3156
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
scott@slp53.sl.home (Scott Lurndal) writes:
> I'm sorry - but it was quite common for IBM customers to
> tweek the OS. Significantly in many cases. I don't doubt that
> many also had custom block-mux channels for various purposes.
re:
http://www.garlic.com/~lynn/2017.html#58 The ICL 2900
that got a lot more difficult with the OCO-wars (ibm stop providing
source). in the wake of various legal actions, ibm has the 23jun1969
"unbundling" announcement ... starting to charge for software,
maintenance, SE services, etc
http://www.garlic.com/~lynn/submain.html#unbundle
.... however, IBM managed to make the case that kernel software should
still be free. during the FS period (was completely different than
360/370 and was completely replace 360/370), 370 efforts were being
shutdown. The lack of 370 offerings during the FS period is credited
with giving clone processor makers market foothold
http://www.garlic.com/~lynn/submain.html#futuresys
when FS imploded, there was a mad rush to get stuff back into the 370
pipelines ... that contributed to decision to release a bunch of 370
stuff I had been doing all during the FS period (I would also
periodically ridicule FS activities, which wasn't exactly career
enhancing).
In the morph of CP67 to VM370 there was lots of stuff dropped that I had
been doing back to time I was undergraduate. I then migrated a bunch of
the stuff to VM370 and one of my hobbies was providing production
enhanced operating systems for internal data centers ... some old email
http://www.garlic.com/~lynn/2006v.html#email731212
http://www.garlic.com/~lynn/2006w.html#email750102
http://www.garlic.com/~lynn/2006w.html#email750430
at the time "csc/vm" was running on couple hundred systems. A lot of my
stuff was picked up for vm370 release 3. However, the rise of clone
processors also motivated the decision to transition to charging for
kernel software ... and bunch of my stuff (kernel reorg for SMP, but not
SMP support itself, redo of paging & virtual memory support, and dynamic
adaptive resource management) was selected as guinea pig
http://www.garlic.com/~lynn/subtopic.html#fairshare
http://www.garlic.com/~lynn/subtopic.html#wsclock
http://www.garlic.com/~lynn/subtopic.html#smp
IBM went thru transition period ... that previous kernel software was
still free, but increasingly amounts were charged for ... until switched
to all kernel software was priced. Then came the "OCO-wars" where source
was no longer shipped. Part of the IBM argument was that customer
changes inhibited moving customers to new versions of software and
hardware models (possibly even software that was increasingly difficult
to run on clone processors).
trivia: TYMSHARE started offering their (vm370) CMS-based online
computer conferencing to (IBM mainframe user group) SHARE for free as
VMSHARE in Aug1976. The VMSHARE archives are here
http://vm.marist.edu/~vmshare/
which includes some amount of customer discussions about the OCO-wars
.... some past posts referencing VMSHARE archive OCO-war discussions.
http://www.garlic.com/~lynn/2004p.html#5 History of C
http://www.garlic.com/~lynn/2004p.html#21 need a firewall
http://www.garlic.com/~lynn/2007f.html#67 The Perfect Computer - 36 bits?
http://www.garlic.com/~lynn/2007k.html#15 Data Areas Manuals to be dropped
http://www.garlic.com/~lynn/2007u.html#8 Open z/Architecture or Not
http://www.garlic.com/~lynn/2007u.html#9 Open z architecture and Linux questions
http://www.garlic.com/~lynn/2008r.html#66 OCO, documentation, support from IBM-Main, etc
http://www.garlic.com/~lynn/2011g.html#29 Congratulations, where was my invite?
http://www.garlic.com/~lynn/2012j.html#20 Operating System, what is it?
http://www.garlic.com/~lynn/2012j.html#30 Can anybody give me a clear idea about Cloud Computing in MAINFRAME ?
http://www.garlic.com/~lynn/2012j.html#31 How smart do you need to be to be really good with Assembler?
http://www.garlic.com/~lynn/2013m.html#55 'Free Unix!': The world-changing proclamation made 30 years ago today
http://www.garlic.com/~lynn/2013o.html#45 the nonsuckage of source, was MS-DOS, was Re: 'Free Unix!
http://www.garlic.com/~lynn/2014.html#19 the suckage of MS-DOS, was Re: 'Free Unix!
http://www.garlic.com/~lynn/2014m.html#35 BBC News - Microsoft fixes '19-year-old' bug with emergency patch
http://www.garlic.com/~lynn/2015d.html#59 Western Union envisioned internet functionality
--
virtualization experience starting Jan1968, online at home since Mar1970
|
|
|
|
| Re: The ICL 2900 [message #336258 is a reply to message #336229] |
Tue, 24 January 2017 13:38 |
Anne & Lynn Wheel
Messages: 3156
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
hancock4@bbs.cpcn.com writes:
> Back then some customers had emotional prejudices which influenced
> their selection of computer. As we know, many customers felt, "you
> never get fired for buying IBM", and went that way, even if the
> price/performance/quality might have been better with alternatives.
> On the other hand, there were some customers who disliked IBM for
> various reasons, and refused to consider an IBM machine.
re:
http://www.garlic.com/~lynn/2017.html#58 The ICL 2900
http://www.garlic.com/~lynn/2017.html#59 The ICL 2900
I had co-worker at IBM Research that left and was doing a lot of
consulting work in silicon valley. For a long time, he had major chip
design customer ... working for the Senior VP for Engineering (that had
started using cp67/cms as young silicon valley engineer in the 60s). He
had done port of AT&T C compiler to vm370/cms with a lot of 370
performance optimizations and ported a lot of UCB chip design tools.
He was then doing ethernet support ... for connecting SGI graphics
terminals to backup 370 "server". The IBM salesman came by and asked him
what he was doing. The IBM salesman then told him that he should do
token-ring support instead ... or he might find that mainframe service
wouldn't be as timely as it had been in the past. I then got an hour
phone call that involved a lot of four letter words. The next morning
the senior VP of Engineering held press conference to announce that they
were moving off IBM mainframes to sun servers.
Then there was a lot of internal task-forces looking at technical
reasons that chip industry was moving off IBM mainframes ... totally
ignoring the salesman issue.
We saw something similar from the communication group with internal
politics (sna, vtam, token-ring, etc) ... recent post mentioning
internal politics preventing us from bidding on NSFNET (precursor to
modern internet).
http://www.garlic.com/~lynn/2017.html#47 Putting The Times's First Email Address to Bed
other past posts
http://www.garlic.com/~lynn/subnetwork.html#nsfnet
also for various reasons the original mainframe TCP/IP product appeared
to be significantly crippled (possibly to make sure that sna/vtam was
faster). I then did the changes to support RFC1044 ... and in some tests
at cray research between 4341 and cray ... demonstrated sustained
channel media throughput using only modest amount of 4341 CPU (possibly
500 times improvement in bytes moved per instruction executed)
http://www.garilc.com/~lynn/subnetwork.html#rfc1044
--
virtualization experience starting Jan1968, online at home since Mar1970
|
|
|
|
| Re: The ICL 2900 [message #336259 is a reply to message #336255] |
Tue, 24 January 2017 14:45 |
Rich Alderson
Messages: 489
Registered: August 2012
Karma: 0
|
Senior Member |
|
|
Anne & Lynn Wheeler <lynn@garlic.com> writes:
> Bob Eager <news0006@eager.cx> writes:
>> Yes, we got a VAXcluster (the base CPU power of an 8800, running SMP, was
>> a bit under the IBM offering, so they chucked in a couple of 8200s). In
>> fact, since it was on VMS 4.6, it couldn't do SMP anyway, and didn't for
>> years.
> old reference to announce of "real" VMS SMP ("DEC Stalks Big Game with
> Symmetrical VMS")
> http://www.garlic.com/~lynn/2007.html#email880329
Which was greeted with hoots of laughter from the (sadly moribund) 36-bit DEC
customers, who had SMP on the PDP-10 architecture nearly a decade earlier. :-(
--
Rich Alderson news@alderson.users.panix.com
Audendum est, et veritas investiganda; quam etiamsi non assequamur,
omnino tamen proprius, quam nunc sumus, ad eam perveniemus.
--Galen
|
|
|
|
| Re: The ICL 2900 [message #336261 is a reply to message #336232] |
Tue, 24 January 2017 16:35 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Monday, January 23, 2017 at 6:09:42 PM UTC-5, g4ugm wrote:
> In general you could get better price/performance elsewhere. IBM's
> technique was to sell to high management, and convince them. The
> technical sales teams were not used to having to work.
In terms of hardware, the other companies generally offered
a better deal in terms of price/performance.
However, in terms of support, IBM generally was superior.
In my personal experience with employers utilizing IBM, Burroughs,
Wang, and Univac hardware, the superior support provided by IBM far
outweighed the superior hardware by the competition. Indeed,
one employer had an IBM on third-party lease (which IBM hated),
so technically we weren't even an IBM customer. None the less,
we still received valuable support from IBM.
We found that IBM employees were far better trained and supported
than employees of the competitors. For hardware issues, spare
parts were readily at hand, even for older machines, and they
quickly brought in experts to fix difficult problems. The
Univac and Burroughs people were really nice folks, but they
just didn't have the training or resources.
In terms of exaggerated claims to make a sale, we found Burroughs,
Univac, and Wang did that plenty, too.
> I remember when the company I worked for was looking at buying a new
> machine around the same time as ICL2900 appeared, IBM suggested that we
> run on-line during the day and batch overnight. Went down like a lead
> balloon with management.
>
> We were Honeywell and bought a Honeywell L66/10 and did run batch and
> on-line at the same time, albeit after a memory upgrade...
>
> Dave
> G4UGM
|
|
|
|
| Re: The ICL 2900 [message #336262 is a reply to message #336250] |
Tue, 24 January 2017 16:37 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Tuesday, January 24, 2017 at 9:43:36 AM UTC-5, jmfbahciv wrote:
> Hardware was always dreadful back then. It was up to the software to
> deal with it and make the system work well despite the hardware
> limitations.
What is "back then" Our S/360-40 generally was extremely reliable
in the late 1970s, when it was technically obsolete.
Our Univac 90/30 was clearly a "lightweight" in terms of durability
and mechanical reliability.
|
|
|
|
| Re: The ICL 2900 [message #336263 is a reply to message #336251] |
Tue, 24 January 2017 16:43 |
hancock4
Messages: 6746
Registered: December 2011
Karma: 0
|
Senior Member |
|
|
On Tuesday, January 24, 2017 at 9:43:36 AM UTC-5, jmfbahciv wrote:
> IBM sales people could be arrogant because they didn't believe they had
> any competition.
In the early years of System/360, IBM had tremendous sales. Likewise
in the postwar era with tabulating machines. However, beyond those
two relatively limited time frames, IBM always had competition in
various ways. Before WW II IBM had to convince customers of the
benefits of automation, despite the high costs of machines, cards,
and skilled operators. After S/360, IBM had to competitors who
offered a better price. Further, IBM had several delays in
delivering S/360 software and hardware, which cost it business and
reputation. By S/370 days, there was plug compatible competition
and then mini-computer competition.
IBM's S/3x series had heavy mini-computer competition. Personally,
I saw a lot more Wang, Data General, and DEC out there than S/3x.
|
|
|
|
| Re: The ICL 2900 [message #336264 is a reply to message #336259] |
Tue, 24 January 2017 17:18 |
Anne & Lynn Wheel
Messages: 3156
Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Rich Alderson <news@alderson.users.panix.com> writes:
> Which was greeted with hoots of laughter from the (sadly moribund) 36-bit DEC
> customers, who had SMP on the PDP-10 architecture nearly a decade earlier. :-(
re:
http://www.garlic.com/~lynn/2017.html#58 The ICL 2900
http://www.garlic.com/~lynn/2017.html#59 The ICL 2900
http://www.garlic.com/~lynn/2017.html#60 The ICL 2900
I periodically mention that Charlie had invented compare&swap when he
was working on CP67 fine-grain kernel multiprocessing locking at the
science center (came up with "compare&swap" because CAS are Charlie's
initials)
http://www.garlic.com/~lynn/subtopic.html#545tech
in the initial morph from CP67 to VM370 lots of my stuff was dropped
(back to my days as undergraduate) ... but also SMP support.
we had a number of projects to put SMP support into VM370 ... thus my
upthread comment about including kernel reorg for SMP ... but not actual
SMP support itself.
however the kernel reorg stuff was included as part of my "charged-for"
resource manager (as well as bunch of my other stuff) ... during the
transition period, older kernel code and hardware support would still be
free ... but could charge for new (non-hardware) support kernel
software. The problem with shipping SMP support in VM370 Release 4 (for
free) was that the kernel reorged needed for SMP support was already
part of my charged for kernel add-on. The eventual resolution was to
moved nearly 90% of the code from my charged for kernel add-on into the
free kernel base (w/o changing the price for my kernel add-on).
http://www.garlic.com/~lynn/subtopic.html#fairshare
http://www.garlic.com/~lynn/subtopic.html#wsclock
also, the initial effort to get compare&swap added to 370 architecture
was rebuffed, the POK favorite son operating system people saying that
test&set (from 360) days was more than adequate. The 370 architecture
owners said that to justify compare&swap for 370 ... uses other than
just kernel SMP locking was required ... thus was born the descriptions
(still in appendix of principles of operation) for use by large
multi-thread applications (like large DBMS transaction systems). In the
80s, you start to see other plaforms implementing compare&swap (or
instructions with similar semantics) for large multi-thread commercial
applications.
http://www.garlic.com/~lynn/subtopic.html#smp
--
virtualization experience starting Jan1968, online at home since Mar1970
|
|
|
|
| Re: The ICL 2900 [message #336265 is a reply to message #336259] |
Tue, 24 January 2017 17:19 |
|
|
Originally posted by: Bob Eager
On Tue, 24 Jan 2017 14:45:25 -0500, Rich Alderson wrote:
> Anne & Lynn Wheeler <lynn@garlic.com> writes:
>
>> Bob Eager <news0006@eager.cx> writes:
>
>>> Yes, we got a VAXcluster (the base CPU power of an 8800, running SMP,
>>> was a bit under the IBM offering, so they chucked in a couple of
>>> 8200s). In fact, since it was on VMS 4.6, it couldn't do SMP anyway,
>>> and didn't for years.
>
>> old reference to announce of "real" VMS SMP ("DEC Stalks Big Game with
>> Symmetrical VMS")
>> http://www.garlic.com/~lynn/2007.html#email880329
>
> Which was greeted with hoots of laughter from the (sadly moribund)
> 36-bit DEC customers, who had SMP on the PDP-10 architecture nearly a
> decade earlier. :-(
We did too. Our 2900 had it on our homebrew operating system. It worked
well, once I'd worked out how to do the inter-CPU control (I had to
reverse engineer the microcode to get a register spec). I didn't writye
the code, but there were model-specific issues.
The guy who wrote *that* SMP code was later hired by DEC (VMS
Engineering). Guess what he worked on.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336266 is a reply to message #336265] |
Tue, 24 January 2017 17:23 |
Quadibloc
Messages: 4399
Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Tuesday, January 24, 2017 at 3:19:27 PM UTC-7, Bob Eager wrote:
> The guy who wrote *that* SMP code was later hired by DEC (VMS
> Engineering). Guess what he worked on.
Hmm. When it comes to the VAX, "steal from the best" seems to have applied both
ways...
John Savard
|
|
|
|
| Re: The ICL 2900 [message #336267 is a reply to message #336266] |
Tue, 24 January 2017 18:01 |
|
|
Originally posted by: Bob Eager
On Tue, 24 Jan 2017 14:23:36 -0800, Quadibloc wrote:
> On Tuesday, January 24, 2017 at 3:19:27 PM UTC-7, Bob Eager wrote:
>
>> The guy who wrote *that* SMP code was later hired by DEC (VMS
>> Engineering). Guess what he worked on.
>
> Hmm. When it comes to the VAX, "steal from the best" seems to have
> applied both ways...
It'll be interesting if I ever get my 2900 emulator to do two CPUs.
(note: they weren't called CPUs; they were called OCPs - Order Code
Processors)
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
| Re: The ICL 2900 [message #336268 is a reply to message #336262] |
Tue, 24 January 2017 18:33 |
Jon Elson
Messages: 646
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
hancock4@bbs.cpcn.com wrote:
> On Tuesday, January 24, 2017 at 9:43:36 AM UTC-5, jmfbahciv wrote:
>
>> Hardware was always dreadful back then. It was up to the software to
>> deal with it and make the system work well despite the hardware
>> limitations.
>
> What is "back then" Our S/360-40 generally was extremely reliable
> in the late 1970s, when it was technically obsolete.
>
> Our Univac 90/30 was clearly a "lightweight" in terms of durability
> and mechanical reliability.
Interesting. I don't know so much about the smaller ones, but we did know
that the larger models (/50 and /65) were pretty sensitive to power
disturbances. The 360/50 at Washington University had an uptime of MAYBE 5
hours. When they upgraded to a 360/65, it seemed to do better. They rented
a Dranetz analyzer and quickly found a correlation between system crashes
and power disturbances. We used to have really ROTTEN power here, due to
crummy underground cables. Too late for the /50 and /65, they finally
discovered shielded cable in the late 1970's and the power problems largely
went away. (The old plain rubber-covered cable was pulled into buried
conduits that leaked groundwater, so the cables with 4160 V AC on them were
partially immered in dirty water, and the corona degraded the rubber. The
cables would flash over for a few weeks before going to catastrophic
failure, and you could see the fluorescent lights flicker all the time.)
Later, the industry moved to shielded cable that had an expanded copper
screen over the main insulation, then covered with a thin outer cover. This
got rid of the electric field on the outside of the cable, and stopped the
corona degradation of the insulation. Some cable somewhere on campus used
to fail every 1-2 months before, I suspect a bunch of the first shielded
cables are still in the ground 40+ years later.
So, anyway, they were able to correlate at least 50% of the system crashes
with a disturbance reported by the Dranetz. The bought a big DPP (Digital
Power Products) system with isolation transformers and breaker panels, but
these could not get rid of actual dropouts in the mains voltages.
The larger 360's, at least, had "converter inverters" that converted 208 V
3-phase mains power to 300 V DC, and then inverted that to 120 V 2500 Hz
sine-wave power, to run all the power supplies around the computer. The
individual supplies had minimal energy storage, and the converter inverter
had a bank of "computer grade" caps in it, but only gave a few milliseconds
of holdup time. So, the system was pretty susceptible to crummy power.
Some other locations that had similar machines didn't seem to suffer this
trouble, as their power was apparently more stable.
370's put in motor generator sets that really isolated the machine from
momentary dips.
Jon
|
|
|
|
| Re: The ICL 2900 [message #336284 is a reply to message #336259] |
Wed, 25 January 2017 08:56 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Rich Alderson wrote:
> Anne & Lynn Wheeler <lynn@garlic.com> writes:
>
>> Bob Eager <news0006@eager.cx> writes:
>
>>> Yes, we got a VAXcluster (the base CPU power of an 8800, running SMP, was
>>> a bit under the IBM offering, so they chucked in a couple of 8200s). In
>>> fact, since it was on VMS 4.6, it couldn't do SMP anyway, and didn't for
>>> years.
>
>> old reference to announce of "real" VMS SMP ("DEC Stalks Big Game with
>> Symmetrical VMS")
>> http://www.garlic.com/~lynn/2007.html#email880329
>
> Which was greeted with hoots of laughter from the (sadly moribund) 36-bit
DEC
> customers, who had SMP on the PDP-10 architecture nearly a decade earlier.
:-(
>
VMS management tried to get Husvedt(sp?) to plan and do SMP on the VAX.
He was sent to JMF and TW to find out about the solutions, problems,
and what JMF/TW would have done if they could do it over. The fucking
idiot got an immediate NIH attitude and left the meeting swearing that
VMS would NEVER have SMP. So that's why it was implemented too late.
/BAH
|
|
|
|
| Re: The ICL 2900 [message #336285 is a reply to message #336252] |
Wed, 25 January 2017 08:56 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
Bob Eager wrote:
> On Tue, 24 Jan 2017 14:43:01 +0000, jmfbahciv wrote:
>
>> Bob Eager wrote:
>>> On Mon, 23 Jan 2017 14:54:44 +0000, Stephen Wolstenholme wrote:
>>>
>>>> In the 1970s I was working on IBM 360 compatible ICL 4/70. Later I
>>>> worked on the ICL 2900 and then it's replacement S39. On the customer
>>>> sites I was working on the ICL systems were faster than IBM systems
>>>> doing the same jobs but the IBM machines were more reliable. An
>>>> interesting thing I noticed was that most customers went for neither
>>>> ICL or IBM when the systems were replaced.
>>>
>>> Reliability (hardware wise) was dreadful. A lot of the stuff in our
>>> operating system was recovery from transient errors.
>>
>> Hardware was always dreadful back then. It was up to the software to
>> deal with it and make the system work well despite the hardware
>> limitations.
>
> Well, no. The reliabity of our DEC stuff was good. And the reliability of
> that machine's immediate predecessor was a lot better too.
You had PDP-10s? Reliability of any hardware was good only if PMs
were religiously done according to FS recommendations. Even then,
the hardware could suck big. Systems should be up and running 24/7
but that was never accomplished back then. The CMs (corrective
maintenance) were usually minimal but only if the PMs (preventive
maintenance) had been done regularly.
Mechanical things and power fluctuations along with environmental
vagaries created many headaches and down times.
/BAH
|
|
|
|
| Re: The ICL 2900 [message #336286 is a reply to message #336263] |
Wed, 25 January 2017 08:56 |
jmfbahciv
Messages: 6173
Registered: March 2012
Karma: 0
|
Senior Member |
|
|
hancock4@bbs.cpcn.com wrote:
> On Tuesday, January 24, 2017 at 9:43:36 AM UTC-5, jmfbahciv wrote:
>
>> IBM sales people could be arrogant because they didn't believe they had
>> any competition.
>
> In the early years of System/360, IBM had tremendous sales. Likewise
> in the postwar era with tabulating machines. However, beyond those
> two relatively limited time frames, IBM always had competition in
> various ways. Before WW II IBM had to convince customers of the
> benefits of automation, despite the high costs of machines, cards,
> and skilled operators. After S/360, IBM had to competitors who
> offered a better price. Further, IBM had several delays in
> delivering S/360 software and hardware, which cost it business and
> reputation. By S/370 days, there was plug compatible competition
> and then mini-computer competition.
>
> IBM's S/3x series had heavy mini-computer competition. Personally,
> I saw a lot more Wang, Data General, and DEC out there than S/3x.
That was later. I was thinking about ( and should have written)
late 60s and early 70s.
/BAH
|
|
|
|
Members
Search
Help
Register
Login
Home
