Old discussion about Windows running on a mainframe ( I brought up)
Ed Gould
Track this topic Print story Post comment
z/VMs get Microsoft rumor
By Timothy Prickett Morgan • Get more from this author
Posted in Servers, 23rd March 2009 22:34 GMT
Whitepaper download - Eight CRM essentials
An obscure mainframe software company called Mantissa Corporation bragged last summer on the IBM VM listserv - which is dedicated to virtual mainframe environments - that it was creating a product called z/VOS that would allow slices of a Windows operating system to run atop z/VM, the hypervisor-as-operating system for IBM mainframes. The product was due in the first quarter of this year, and the story of its impending release has been making the rounds.
According to a report in NetworkWorld, Mantissa's z/VOS, presumably short for Virtual Operating System, is a layer of software for VM that allows desktop and server Windows operating systems to run in emulated mode atop z/VM. Mantissa - which is based in Birmingham, Alabama, and which is a supplier of report distribution and other tools for mainframes - talked about the z/VOS product at the SHARE mainframe user conference in early March in Austin, Texas. But that was not the same thing as a product launch.
We've tried to reach the company for several days, but Mantissa has yet to respond.
While IBM and the Linux community for mainframes centered around Marist College in New York have worked to get official mainframe ports done for Linux - Red Hat and Novell officially support mainframes, if you can write a big enough check to get support - there is no native Windows port to IBM mainframes as far as I know. So, the real curiosity is how Mantissa is supporting Windows XP or Vista atop z/VM partitions.
According to the company's development blog, z/VOS includes a translation engine that "converts native x86 code to its System z equivalent." See how easy that was? As it translates equivalent results - not creating equivalent machine code, mind you - the instruction that is created by z/VOS is stored in memory so it can be accessed the next time the operating system function inside Windows running on the mainframe is asked for again.
Since Gary Dennis, Mantissa's chief executive officer and founder - and other we've called - are not answering their phones, it is a little hard to take the company seriously. But if it can indeed deliver a layer of abstraction software atop z/VM that lets Windows desktops and servers run on mainframe iron, the company should probably think about getting someone to answer the phones and maybe a salesperson or two to try to take some orders. If the x86 translation overhead is not too high, this could be a very interesting development - and one that Big Blue would seem pretty keen on supporting, not quashing. ®
http://www.theregister.co.uk/2009/03/23/mantissa_windows_on_mainframes/
AT least these people aren't IBM types and IBM is not eager to dispel it (so I have heard)
Ed
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to list...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
Harry Wahl
Many people accuse IBM of acquiring PSI to stifle competition, but this is a great disservice to the engineers at PSI, particularly former Amdahl engineers, who are now bound to silence by NDAs.
IBM may see the virtualization of Windows onto z architecture as strategic; both as servers and workstations and its acquisition of PSI reflects this.
From what little Mantissa has disclosed, their approach, IMHO, is inherently flawed. Their SHARE presentation seemed to be little more than an obfuscated discourse on virtualization in general. Even if their approach works, it would have been better, architecturally, to emulate an Itanium for Windows hosting because of the nature of the Itanium’s instruction parallelism.
A more sensible approach would be to look at creating a Windows HAL (hardware abstraction layer), or something conceptually similar, that runs on z/Series. Historically, this is how Windows has been made to run on different machine architectures. Of course, cooperative development between IBM and Microsoft would be necessary.
Another possibility is to exploit the Infiniband feature of the z/10. This feature is profound in terms of 360-z/series evolution, but has been largely ignored, so far.
Infiniband attached external hardware products that expose x86 architecture processors, from Intel or IBM itself (or maybe even Intel Larrabee), is an ideal way to run Windows on IBM mainframes.
In the interest of disclosure, I have worked for both IBM and PSI, but the opinions I express here are complete conjecture.
Harry
Harry...@Hotmail.com
> Date: Mon, 23 Mar 2009 21:08:01 -0700
> From: ps2...@yahoo.com
> Subject: Old discussion about Windows running on a mainframe ( I brought up)
> To: IBM-...@bama.ua.edu
Kirk Wolf
> Platform Solutions, Inc. had this technology and much more. Their
> technology is now IBM's.
>
>
Doesn't PSI does the opposite of what Mantissa claims to be working on?
PSI runs z architecture on x86 hardware whereas Mantissa claims to run x86
architecture on z hardware.
Combine the two and you could run Windows on z/VM on a x86 emulating a z
machine. Sounds snappy :-)
BTW: MacOS did this quite successfully when they switched from 68000 to
Power architecture. You could run your 68000 binaries and they would be
dynamically translated into a cache Power instructions. They did it again
when they moved to x86.
See: http://knowledgerush.com/kr/encyclopedia/Mac_68K_emulator/
Of course, modern byte-code VMs use dynamic translation ("JIT") to translate
frequently used byte codes into native machine instructions. Yet, it is
still amazing how many people think that Java is still only "interpreted".
Quoting from an Austin SHARE presentation on PHP:
"Unlike Java or other interpretive language, PHP’s focus is to
use the script to invoke “native” C subroutines that do the
actual work at full speed"
Kirk Wolf
Dovetailed Technologies
John McKown
Java JVM. The z/VOS reads and interprets the x86 arch instructions, with
some JIT on the side. However, other companies, such as AMD, need cross
licensing to create chips which run x86. That "cross licensing" is part of
IBM's complaint against PSI. PSI was not licensed to create a zArchitecture
emulator. Couldn't Intel "shut down" Mantissa with a similar suite? I don't
remember reading that Mantissa has an agreement with Intel. I don't think
the "fact" that z/VOS is a software emulator instead of a harware
implementation is relevant. Of course, IANAL.
--
John
P S
Emulation *might* be different--PSI kept saying they didn't do
emulation, although it quacked like a duck.
The AMD licensing might be slightly different, since their
architecture is slightly different.
But it's a great question! I'd love to know the answer.
P.S. "z/Architecture". Slash for software (and the architecture, oddly
enough, is considered software).
On Tue, Mar 24, 2009 at 12:36 PM, John McKown <joa...@swbell.net> wrote:
> I'm a bit curious. If I understand what is being done, it is similar to a
> Java JVM. The z/VOS reads and interprets the x86 arch instructions, with
> some JIT on the side. However, other companies, such as AMD, need cross
> licensing to create chips which run x86. That "cross licensing" is part of
> IBM's complaint against PSI. PSI was not licensed to create a zArchitecture
> emulator. Couldn't Intel "shut down" Mantissa with a similar suite? I don't
> remember reading that Mantissa has an agreement with Intel. I don't think
> the "fact" that z/VOS is a software emulator instead of a harware
> implementation is relevant. Of course, IANAL.
----------------------------------------------------------------------
Gibney, Dave
> From: IBM Mainframe Discussion List [mailto:IBM-...@bama.ua.edu] On
> Behalf Of Harry Wahl
> Sent: Tuesday, March 24, 2009 5:35 AM
> To: IBM-...@bama.ua.edu
> Subject: Re: Old discussion about Windows running on a mainframe ( I
> brought up)
>
>
>
> (hardware abstraction layer), or something conceptually similar, that
> runs on z/Series. Historically, this is how Windows has been made to
> run on different machine architectures. Of course, cooperative
> development between IBM and Microsoft would be necessary.
>
>
>
> Another possibility is to exploit the Infiniband feature of the z/10.
> This feature is profound in terms of 360-z/series evolution, but has
> been largely ignored, so far.
>
>
>
> Infiniband attached external hardware products that expose x86
> architecture processors, from Intel or IBM itself (or maybe even Intel
> Larrabee), is an ideal way to run Windows on IBM mainframes.
>
>
>
>
It seems to me that the obvious way to do this would be to build z86
on z in millicode. Then, the only real issue would be performance. An
all millicode representation would be "somewhat" slower than the
silicon.
Dave Gibney
Information Technology Services
Washington State University
Tom Marchant
>I'm a bit curious. If I understand what is being done, it is similar to a
>Java JVM. The z/VOS reads and interprets the x86 arch instructions, with
>some JIT on the side. However, other companies, such as AMD, need cross
>licensing to create chips which run x86.
Do they?
>That "cross licensing" is part of
>IBM's complaint against PSI. PSI was not licensed to create a zArchitecture
>emulator.
I thought that the problem with PSI (and Hercules, for that matter) was not
about the emulation of the hardware, but the licensing of z/OS.
--
Tom Marchant
John McKown
wrote:
>On Tue, 24 Mar 2009 11:36:42 -0500, John McKown wrote:
>
>>I'm a bit curious. If I understand what is being done, it is similar to a
>>Java JVM. The z/VOS reads and interprets the x86 arch instructions, with
>>some JIT on the side. However, other companies, such as AMD, need cross
>>licensing to create chips which run x86.
>
>Do they?
I am certain that AMD and Via have cross licensing agreements with Intel. I
know because Intel has recently indicated that AMD is violating that agreement.
http://www.amd.com/us-en/Weblets/0,,7832_12670_12686,00.html
>
>>That "cross licensing" is part of
>>IBM's complaint against PSI. PSI was not licensed to create a zArchitecture
>>emulator.
>
>I thought that the problem with PSI (and Hercules, for that matter) was not
>about the emulation of the hardware, but the licensing of z/OS.
Hum, parts of z/Architecture are patented. Not the hardware to perform the
function, but the actual function itself - indepandant of the
implementation. Neither Hercules/390 nor PSI had a license for those parts
of the z/Architecture, even though they implemented them (not totally sure
about Hercules/390 on this point).
Also, as I remember (Danger, Will Robinson!), part of IBM's argument was
that PSI's implementation of the z/Architecture was not verified by IBM. IBM
said that in that case, the use of licensed IBM software (z/OS et al.) could
not be guaranteed. However, much of IBM's "premium" for their software was
due to IBM's guarantees on the reliability of the software. If the hardware
was "not up to snuff", then IBM might suffer damage to its reputation for
reliability if the software (z/OS) were to fail. This was the "logic" as to
why IBM refused to license z/OS et al. on the PSI solution.
>
>--
>Tom Marchant
--
John
Tom Marchant
>On Tue, 24 Mar 2009 12:49:26 -0500, Tom Marchant wrote:
>
>>On Tue, 24 Mar 2009 11:36:42 -0500, John McKown wrote:
>>
>>>.... However, other companies, such as AMD, need cross
>>>licensing to create chips which run x86.
>>
>>Do they?
>
>I am certain that AMD and Via have cross licensing agreements with Intel. I
>know because Intel has recently indicated that AMD is violating that agreement.
>
>http://www.amd.com/us-en/Weblets/0,,7832_12670_12686,00.html
The only thing I see in the page like that is this:
"In October 1991, Intel commenced a federal court action for copyright
infringement. An arbitrator subsequently awarded AMD full rights to make and
sell the Am386. The Supreme Court of California upheld this decision in 1994."
Is that what you are talking about?
>
>>
>>>That "cross licensing" is part of
>>>IBM's complaint against PSI. PSI was not licensed to create a zArchitecture
>>>emulator.
>>
>>I thought that the problem with PSI (and Hercules, for that matter) was not
>>about the emulation of the hardware, but the licensing of z/OS.
>
>
>Also, as I remember (Danger, Will Robinson!), part of IBM's argument was
>that PSI's implementation of the z/Architecture was not verified by IBM. IBM
>said that in that case, the use of licensed IBM software (z/OS et al.) could
>not be guaranteed. However, much of IBM's "premium" for their software was
>due to IBM's guarantees on the reliability of the software. If the hardware
>was "not up to snuff", then IBM might suffer damage to its reputation for
>reliability if the software (z/OS) were to fail. This was the "logic" as to
>why IBM refused to license z/OS et al. on the PSI solution.
And one of the factors that makes the 360/370/z et. al. reliable is the
machine check interruption. Does the Itanium processor have that
capability? I'm pretty sure that none of the x86 processors do.
--
Tom Marchant