AMD to leave x86 behind?

David Brown wrote:

Speaking as someone who works with small embedded systems, that's simply
not true. ARM is certainly becoming more popular, and taking over
markets previously occupied by high-end 16-bit microcontrollers, and
occasionally high-end 8-bitters. ARM is competing with other 32-bit
architectures, such as Coldfire, which are far from being on the way
out. There are continuous introductions of new 8-bit chips from big
players like Freescale (né Motorola), Atmel, and Microchip - often
updated and modernized cores rather than the originals. About the only
popular small core that does not have major backing from big players is
the 8051. It is popular among small, specialised chip manufacturers,
and will continue for years as standard parts, but is being dropped in
new devices in favour of other cores that are cheaper and more efficient.

There's no doubt of ARM's increasing popularity, but rumours of its
impending monopoly of embedded systems are greatly exaggerated.

Have the older Intel RISC chips gone away then? From memory i860 and
i960, one of which had an exposed pipeline. How soon and happily they
forget ;-)

--
bill davidsen
SBC/Prodigy Yorktown Heights NY data center
http://newsgroups.news.prodigy.com

Tony Hill wrote:
On Mon, 21 Nov 2005 22:19:09 GMT, Mike Hore
wrote:


Jeremy Linton wrote:



As for instruction sets though, I stick by my thoughts that there are
only 3 instructions sets with much of a futu x86, PowerPC and ARM.

I would add 8051 derivatives to that list.

And don't write off the 360/370/390/z/Architecture just yet -- it's
stuck around much longer than anybody thought in 1964.


Note that I said a *FUTURE*. I'm not saying that all other
architectures will disappear off the face of the planet. S390 will
still exist for some time to come, but lets face it, it's been about
15 years since much of anyone has done significant new development on
mainframes. They exist almost solely to support legacy applications
that are too difficult/expensive to port to a newer design.

I'm sure that others will educate you on this, but Linux on mainframe is
pretty non-legacy, and is being adopted by lots of very non-tech
companies. The Boscov's Department Store chain was the the biggest
surprise to me, Hanaford Markets (NY & New England)went Linux, but I
don't know if that's on mainframe or not, although I suspect it is.

--
bill davidsen
SBC/Prodigy Yorktown Heights NY data center
http://newsgroups.news.prodigy.com

Tony Hill wrote:

I would tend to think that it is, but fortunately for IBM there are
sufficient stupid companies that have decided their old applications
are too expensive to port to new platforms that they continue to
require new mainframe hardware and (more importantly) support
services.

Seriously though, what can you get from a mainframe that you could not
get from 1/4 of the cost, either in PC hardware or in high-end servers
of the IBM Power5 variety. I/O and reliability have been sufficiently
solved for 99.999% of all applications, so what's left? There are
certain security features that mainframes have which PCs may lack, but
history has proven rather nicely that *NO* system is 100% secure and
perhaps the biggest advantage for mainframes in this regard today is
the old "security through obscurity" (or perhaps more
knowledgeable/better trained maintainers!).

I have never seen a new application in the past ~10 years where I
thought that a mainframe would be the appropriate solution.

Tony, you usually know what you're talking about. This is not one of
those times.

--
bill davidsen
SBC/Prodigy Yorktown Heights NY data center
http://newsgroups.news.prodigy.com

"Bill Davidsen" wrote in message
m...
David Brown wrote:

Speaking as someone who works with small embedded systems, that's simply
not true. ARM is certainly becoming more popular, and taking over
markets previously occupied by high-end 16-bit microcontrollers, and
occasionally high-end 8-bitters. ARM is competing with other 32-bit
architectures, such as Coldfire, which are far from being on the way out.
There are continuous introductions of new 8-bit chips from big players
like Freescale (né Motorola), Atmel, and Microchip - often updated and
modernized cores rather than the originals. About the only popular small
core that does not have major backing from big players is the 8051. It
is popular among small, specialised chip manufacturers, and will continue
for years as standard parts, but is being dropped in new devices in
favour of other cores that are cheaper and more efficient.

There's no doubt of ARM's increasing popularity, but rumours of its
impending monopoly of embedded systems are greatly exaggerated.

Have the older Intel RISC chips gone away then? From memory i860 and i960,
one of which had an exposed pipeline. How soon and happily they forget ;-)

Dennis O'Conner can provide better answers than I can, but I believe that
i860 (the one with the exposed pipeline) is long gone, and i960 (which I
wouldn't call RISC) is still (barely) around but with essentially no new
development having been replaced in Intel's product line by Strong ARM and
its follow-ons.

--
- Stephen Fuld
e-mail address disguised to prevent spam

On Tue, 29 Nov 2005 17:33:41 GMT, Bill Davidsen
wrote:
Tony Hill wrote:

Note that I said a *FUTURE*. I'm not saying that all other
architectures will disappear off the face of the planet. S390 will
still exist for some time to come, but lets face it, it's been about
15 years since much of anyone has done significant new development on
mainframes. They exist almost solely to support legacy applications
that are too difficult/expensive to port to a newer design.

I'm sure that others will educate you on this, but Linux on mainframe is
pretty non-legacy, and is being adopted by lots of very non-tech
companies. The Boscov's Department Store chain was the the biggest
surprise to me, Hanaford Markets (NY & New England)went Linux, but I
don't know if that's on mainframe or not, although I suspect it is.

A quick search on Google seems to indicate that Hannaford's big Linux
success story is based around their Point of Sales machines running
the open source OS, not mainframes. In fact, I couldn't find anything
at all about them using mainframes.

As for Boscov, their decision to run Linux on a mainframe seems to be,
in part, related to the fact that they needed a new mainframe anyone
for legacy apps, and rather than have the thing sitting around
twiddling it's thumbs most of the time they decided to put Linux on
the thing and run some of their other apps there. Here's a telling
quote:

"We went with the z900 for additional zOS processing capability, but
we also knew that a portion of the box could be reserved for zVM and
Linux instances. We ordered the machine with the Integrated Facility
for Linux (IFL) right from the factory, and had our first instance
running the same year."

Source:
http://www.linuxplanet.com/linuxplanet/reports/4990/1/


-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

"Tony Hill" wrote in message
...
On Tue, 29 Nov 2005 17:33:41 GMT, Bill Davidsen

A quick search on Google seems to indicate that Hannaford's big Linux
success story is based around their Point of Sales machines running
the open source OS, not mainframes. In fact, I couldn't find anything
at all about them using mainframes.


Considering that there are likely more than 1000 mainframes running Linux,
and the z800 runs only Linux, perhaps you better do a bit more Googling if
you actually want to learn. Otherwise, it only sounds like you want to dig
in your heels and be 'right'.

Regards,
Dean

On Wed, 30 Nov 2005 03:49:22 -0800, "Dean Kent"
wrote:

"Tony Hill" wrote in message
.. .
On Tue, 29 Nov 2005 17:33:41 GMT, Bill Davidsen

A quick search on Google seems to indicate that Hannaford's big Linux
success story is based around their Point of Sales machines running
the open source OS, not mainframes. In fact, I couldn't find anything
at all about them using mainframes.


Considering that there are likely more than 1000 mainframes running Linux,
and the z800 runs only Linux, perhaps you better do a bit more Googling if
you actually want to learn. Otherwise, it only sounds like you want to dig
in your heels and be 'right'.

I'm not the only one digging in my heels :

I did indeed see that there are over 1000 mainframes running Linux.
In fact, IBM sold it's 1000th Linux mainframe back in 2001.

However the flip side to that is that the total MONTHLY volume of
servers sold these days is roughly 575,000 - 600,000. Even given the
HUGE disparity between the cost and capabilities of mainframes vs.
many other servers (note that the number above includes the likes of
IBM's Power servers, Sun SPARC, etc.), I find it tough to look at
mainframes as being anything resembling "the way of the future".

Who knows though, maybe I'm looking at this all from the wrong angle,
but from my point of view, the primary purpose for ever buying a
mainframe is to support legacy apps. Anything else is a *small* niche
or a secondary target.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

"Tony Hill" wrote in message
...

I'm not the only one digging in my heels :

I did indeed see that there are over 1000 mainframes running Linux.
In fact, IBM sold it's 1000th Linux mainframe back in 2001.

However the flip side to that is that the total MONTHLY volume of
servers sold these days is roughly 575,000 - 600,000. Even given the
HUGE disparity between the cost and capabilities of mainframes vs.
many other servers (note that the number above includes the likes of
IBM's Power servers, Sun SPARC, etc.), I find it tough to look at
mainframes as being anything resembling "the way of the future".

I would never claim mainframes are the 'way of the future' - but I would
also not claim Power, x86, ARM or anything else is either because each of
those has it's particular market and will eventually be supplanted by
something else. I posted some mainframe numbers earlier. In the mid 1990's
at the *peak* of the mainframe era, there were between 13K and 14K
mainframes installed. Several years ago it was about 9K. Today there are
between 10K and 11K. More than 10% of those are running Linux. Therefore,
there is growth and much of that has nothing to do with legacy apps.

Now, consider that 50% of IBMs revenues are mainframe related (hardware,
software, services). 50% of CA's revenues are mainframe related. There are
still several other significant mainframe software vendors (Innovation,
Compuware, BMC, Serena) and a number of smaller ones. From what I can
tell, this is a tens of billions of dollars per year industry.


Who knows though, maybe I'm looking at this all from the wrong angle,
but from my point of view, the primary purpose for ever buying a
mainframe is to support legacy apps. Anything else is a *small* niche
or a secondary target.

Unfortunately, there are quite a number of Fortune 1000 CEOs and CIOs that
seem to have a different opinion. What they seem to have realized is that
there is a wide variety of problems to solve, and each has its own solution.
One size does not fit all - not even x86.

Regards,
Dean


-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

In article ,
says...

snip

Now, consider that 50% of IBMs revenues are mainframe related (hardware,
software, services). 50% of CA's revenues are mainframe related. There are
still several other significant mainframe software vendors (Innovation,
Compuware, BMC, Serena) and a number of smaller ones. From what I can
tell, this is a tens of billions of dollars per year industry.

Hmm, taking your assertion that "50% of IBMs revenues are mainframe
related" and IBM's $94B revenue, that's "tens of billions per
year" right there. ;-)

--
Keith

"Keith" wrote in message
...
In article ,
says...

snip

Now, consider that 50% of IBMs revenues are mainframe related (hardware,
software, services). 50% of CA's revenues are mainframe related. There
are
still several other significant mainframe software vendors (Innovation,
Compuware, BMC, Serena) and a number of smaller ones. From what I can
tell, this is a tens of billions of dollars per year industry.

Hmm, taking your assertion that "50% of IBMs revenues are mainframe
related" and IBM's $94B revenue, that's "tens of billions per
year" right there. ;-)

Ah, someone got it. :-).

I also forgot to mention one other interesting set of numbers. Using zVM,
a mainframe can run dozens, hundreds, or perhaps even a thousand or more
Linux images (this is the reason zVM is used rather than hardware
partitioning, which limits you to 15 images). Using the best case of 1000
images, if we have just 1000 mainframes running Linux under zVM, that makes
1,000,000 Linux images - which is a little bit more difficult to sneeze at.

Of course, the reality is that most installations are likely running dozens
of images, and some will be running a hundred or more. That is still tens
of thousands of Linux images on those paltry, legacy only mainframes...
Yep, just a small niche. (grin).

Regards,
Dean


--
Keith