20 Nov
2017
20 Nov
'17
4:05 p.m.
From: Clem Cole <clemc(a)ccc.com>
IIRC Tom Lyons started a 370 port at Princeton and
finished it at
Amdahl. But I think that was using VM
Maybe this is my lack of knowledge of VM showing, but how did having VM help
you over running on the bare hardware?
Noel
20 Nov
20 Nov
4:37 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 11:05 AM, Noel Chiappa <jnc(a)mercury.lcs.mit.edu>
wrote:
Maybe this is my lack of knowledge of VM showing, but how did having VM
help
you over running on the bare hardware?
As an IBM person, I would ask Charlie to answer here, but I believe the
answer from the Locus side was tools primarily and I also think they did
not have to support as much specific HW (*i.e.* smaller foot print of
devices).
4:47 p.m.
New subject: UNIX on S/370
I was waiting for Clem to weigh in on this, since I assume he knows more about it than I
do.
I wasn’t paying much attention to Unix on 370, but my impression has always been that
there were multiple 370 ports. The only ones that were completed, to my knowledge, were
the ESS one and AIX/370. I don’t know of the ESS one being available outside of AT&T.
I don’t know anything about the compilers used, would assume they were PCC-based, even if
provided by IBM.
In 1989, when I left IBM, there were certainly plenty of 370 people inside IBM that would
have understood 370 channels. I would have thought that to still be true in 1991, even
with the buyout packages that encouraged people to retire. (It wasn’t until Gerstner
became CEO in April 1993 that IBM abandoned the “full employment” traditions.)
What Clem says about the smaller footprint of devices rings true. There were also likely
customers that wanted to run AIX alongside other VMs, just as there were customers who
wanted to run MVS alongside other VMs.
Charlie
From: Clem Cole
Sent: Monday, November 20, 2017 10:37 AM
To: Noel Chiappa
Cc: The Eunuchs Hysterical Society
Subject: Re: [TUHS] UNIX on S/370
On Mon, Nov 20, 2017 at 11:05 AM, Noel Chiappa <jnc(a)mercury.lcs.mit.edu> wrote:
Maybe this is my lack of knowledge of VM showing, but how did having VM help
you over running on the bare hardware?
As an IBM person, I would ask Charlie to answer here, but I believe the answer from the
Locus side was tools primarily and I also think they did not have to support as much
specific HW (i.e. smaller foot print of devices).
5:44 p.m.
New subject: UNIX on S/370
On 2017-11-20 11:47 AM, Charles H Sauer wrote:
I was waiting for Clem to weigh in on this, since I
assume he knows more
about it than I do.
I wasn’t paying much attention to Unix on 370, but my impression has
always been that there were multiple 370 ports. The only ones that were
completed, to my knowledge, were the ESS one and AIX/370. I don’t know
of the ESS one being available outside of AT&T.
I don’t know anything about the compilers used, would assume they were
PCC-based, even if provided by IBM.
Yes, Johnson's paper[1] lists pcc's targets as Honeywell 6000, IBM 370,
Interdata 8/32, DG Nova, "and others".
From time to time I wonder what became of those pcc
versions...
--Toby
[1] https://dl.acm.org/citation.cfm?doid=512760.512771
In 1989, when I left IBM, there were certainly plenty of 370 people
inside IBM that would have understood 370 channels. ...
Charlie
*From:* Clem Cole
*Sent:* Monday, November 20, 2017 10:37 AM
*To:* Noel Chiappa
*Cc:* The Eunuchs Hysterical Society
*Subject:* Re: [TUHS] UNIX on S/370
On Mon, Nov 20, 2017 at 11:05 AM, Noel Chiappa <jnc(a)mercury.lcs.mit.edu>
wrote:
Maybe this is my lack of knowledge of VM showing, but how did having
VM help
you over running on the bare hardware?
As an IBM person, I would ask Charlie to answer here, but I believe the
answer from the Locus side was tools primarily and I also think they
did not have to support as much specific HW (/i.e./ smaller foot print
of devices).
7:07 p.m.
New subject: UNIX on S/370
On 11/20/17, Noel Chiappa <jnc(a)mercury.lcs.mit.edu> wrote:
Maybe this is my lack of knowledge of VM showing, but how did having VM
help
you over running on the bare hardware?
It would mean that you wouldn't have to implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
-Paul W.
7:10 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul Winalski wrote:
It would mean that you wouldn't have to implement
machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
Indeed. I'm currently trying to convince Netflix that the way to get the
most performance out of a NUMA machine is to boot a different kernel on
each NUMA domain. One way we might demo that is on a 4 domain system
lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
and give the host kernel the 4th.
7:36 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 12:10 PM, Larry McVoy <lm(a)mcvoy.com> wrote:
On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul
Winalski wrote:
Having a single nic presents a bit of a challenge for this... I look
forward to this demo...
Warner
It would mean that you wouldn't have to
implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
Indeed. I'm currently trying to convince Netflix that the way to get the
most performance out of a NUMA machine is to boot a different kernel on
each NUMA domain. One way we might demo that is on a 4 domain system
lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
and give the host kernel the 4th.
7:36 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 12:36:05PM -0700, Warner Losh wrote:
On Mon, Nov 20, 2017 at 12:10 PM, Larry McVoy
<lm(a)mcvoy.com> wrote:
Nope, do disks/mics/mem per domain. So no bonding.
On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul
Winalski wrote:
Having a single nic presents a bit of a challenge for this... I look
forward to this demo... It would mean that you wouldn't have to
implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
Indeed. I'm currently trying to convince Netflix that the way to get the
most performance out of a NUMA machine is to boot a different kernel on
each NUMA domain. One way we might demo that is on a 4 domain system
lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
and give the host kernel the 4th.
7:39 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 11:36:51AM -0800, Larry McVoy wrote:
On Mon, Nov 20, 2017 at 12:36:05PM -0700, Warner Losh
wrote:
Sorry s/mics/nics/
--
---
Larry McVoy lm at mcvoy.com http://www.mcvoy.com/lm
On Mon, Nov 20, 2017 at 12:10 PM, Larry McVoy
<lm(a)mcvoy.com> wrote:
Nope, do disks/mics/mem per domain. So no bonding. On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul
Winalski wrote:
Having a single nic presents a bit of a challenge for this... I look
forward to this demo... It would mean that you wouldn't have to
implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
Indeed. I'm currently trying to convince Netflix that the way to get the
most performance out of a NUMA machine is to boot a different kernel on
each NUMA domain. One way we might demo that is on a 4 domain system
lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
and give the host kernel the 4th. 
7:56 p.m.
New subject: UNIX on S/370
I would love to see the results of that, including more information
about the architecture in question.
On 11/20/2017 2:10 PM, Larry McVoy wrote:
On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul
Winalski wrote:
It would mean that you wouldn't have to
implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
Indeed. I'm currently trying to convince Netflix that the way
to get the
most performance out of a NUMA machine is to boot a different kernel on
each NUMA domain. One way we might demo that is on a 4 domain system
lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
and give the host kernel the 4th.
7:59 p.m.
New subject: UNIX on S/370
Who knows if it will go anywhere. I got dragged out of retirement with
hints of piles of money, so far, they loaned me a box.
I believe the likely target for this would be AMD's Epyc. They have already
pushed one box to serve up about 100Gbit/sec of movies and that's with them
doing TLS in the kernel; be faster if they could get the NIC to do that.
https://news.ycombinator.com/item?id=15367421
The limiting factor, so far as I can tell, is memory BW and PCIe lanes.
Epyc seems to deliver more of that.
On Mon, Nov 20, 2017 at 02:56:04PM -0500, Arthur Krewat wrote:
I would love to see the results of that, including
more information about
the architecture in question.
On 11/20/2017 2:10 PM, Larry McVoy wrote:
>On Mon, Nov 20, 2017 at 02:07:27PM -0500, Paul Winalski wrote:
>>It would mean that you wouldn't have to implement machine check
>>support and other hardware error handling. The VM hypervisor would do
>>that for you. It would also let you run multiple versions of UNIX
>>simultaneously. Very convenient if you're doing kernel or driver
>>development.
>Indeed. I'm currently trying to convince Netflix that the way to get the
>most performance out of a NUMA machine is to boot a different kernel on
>each NUMA domain. One way we might demo that is on a 4 domain system
>lock down 3 hypervisors and their guest OS to 3/4 of the NUMA domains
>and give the host kernel the 4th.
>
--
---
Larry McVoy lm at mcvoy.com http://www.mcvoy.com/lm
7:27 p.m.
New subject: UNIX on S/370
Paul Winalski <paul.winalski(a)gmail.com> wrote:
On 11/20/17, Noel Chiappa
<jnc(a)mercury.lcs.mit.edu> wrote:
The "simpler hardware handling" is a big inducement to building
on top of VM. Or at least starting out that way.
Even though AIX/370 was aimed at the educational market, my impression
is that it was still pretty expensive. When I worked at the Emory U
computing center (mid 80s) I heard about it, but management wasn't
interested in trying to get it for their S/390.
I'm pretty sure I remember hearing at some point, I don't remember when,
that AIX/370 could run either under VM or on bare iron.
Any idea what Linux/370 does? I think it runs on bare iron.
Arnold
Maybe this is my lack of knowledge of VM showing, but how did having VM
help
you over running on the bare hardware?
It would mean that you wouldn't have to implement machine check
support and other hardware error handling. The VM hypervisor would do
that for you. It would also let you run multiple versions of UNIX
simultaneously. Very convenient if you're doing kernel or driver
development.
7:29 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 12:27:25PM -0700, arnold(a)skeeve.com wrote:
Any idea what Linux/370 does? I think it runs on bare
iron.
Pretty sure that's correct, we used to run Linux/370 on simulator to
build BitKeeper images for it. I don't remember any VM stuff.
10:56 p.m.
New subject: UNIX on S/370
On Mon, 20 Nov 2017, Larry McVoy wrote:
On Mon, Nov 20, 2017 at 12:27:25PM -0700,
arnold(a)skeeve.com wrote:
When I was at IBM, I had a few customers running the product I supported
under Linux on z, and what wikipedia says matches my memory of it being
able to run either on metal or as a VM:
Linux runs on standard, general purpose mainframe CPs (Central
Processors) as well as IFLs (Integrated Facility for Linux). IFLs are
mainframe processors dedicated to running Linux, either natively or
under a hypervisor (z/VM or KVM on z). Microcode restricts IFLs from
running "traditional" workloads, such as z/OS, but they are physically
identical to other z System processors. IFLs are typically less
expensive to acquire from IBM than CPs.
--
Michael Parson
Pflugerville, TX
KF5LGQ
Any idea what Linux/370 does? I think it runs on
bare iron.
Pretty sure that's correct, we used to run Linux/370 on simulator to
build BitKeeper images for it. I don't remember any VM stuff.
11:23 p.m.
New subject: UNIX on S/370
On Mon, Nov 20, 2017 at 04:56:27PM -0600, Michael Parson wrote:
On Mon, 20 Nov 2017, Larry McVoy wrote:
Sorry, we were running Linux/370 on Hercules simulator:
http://www.hercules-390.org/
which I _thought_ emulated the bare metal.
On Mon, Nov 20, 2017 at 12:27:25PM -0700,
arnold(a)skeeve.com wrote:
When I was at IBM, I had a few customers running the product I supported
under Linux on z, and what wikipedia says matches my memory of it being
able to run either on metal or as a VM:
Linux runs on standard, general purpose mainframe CPs (Central
Processors) as well as IFLs (Integrated Facility for Linux). IFLs are
mainframe processors dedicated to running Linux, either natively or
under a hypervisor (z/VM or KVM on z). Microcode restricts IFLs from
running "traditional" workloads, such as z/OS, but they are physically
identical to other z System processors. IFLs are typically less
expensive to acquire from IBM than CPs. Any idea what Linux/370 does? I think it runs on
bare iron.
Pretty sure that's correct, we used to run Linux/370 on simulator to
build BitKeeper images for it. I don't remember any VM stuff.
2354
days inactive
2354
days old
14 comments
10 participants
participants (10)
-
arnold＠skeeve.com -
Arthur Krewat -
Charles H Sauer -
Clem Cole -
jnc＠mercury.lcs.mit.edu -
Larry McVoy -
Michael Parson -
Paul Winalski -
Toby Thain -
Warner Losh