Re: Mainframe running 1,500 Linux servers?
Anne & Lynn Wheeler
> Linux on System z supports Discontinguous Saved Segments (DCSS), a z/VM
> shared memory feature, and has for quite some time. There is a Linux block
> device driver for DCSS (dcssblk.c) in the mainline Linux kernel
> distribution. Features such as shared guest kernel(s), execute-in-place,
> and certain monitoring functions take advantage of this capability, so it's
> quite commonly used. Here's one source of information, for example:
>
> http://www.redbooks.ibm.com/abstracts/sg247285.html
>
> DCSS can be read/write (type SN) or read-only (type SR).
re:
http://www.garlic.com/~lynn/2009q.html#15 Mainframe running 1,500 Linux servers?
I had done the original of DCSS as part of generalized paged mapped
filesystem for CMS. Only a very small read-only subset of that was
shipped in vm370 release 3. Old email discussing migrating the changes
from cp67 to vm370 (early release 3).
http://www.garlic.com/~lynn/2006v.html#email731212
http://www.garlic.com/~lynn/2006w.html#email750102
http://www.garlic.com/~lynn/2006w.html#email750430
none of the paged-mapped filesystem support was shipped (some A/B tests
with modereate i/o thruput cms applications with optimized normal
filesystem on 3380s ... ran avg. 3-times faster with the paged mapped
filesystem ... some operations were significantly faster). misc. past
posts
http://www.garlic.com/~lynn/submain.html#mmap
Part of demise of future system effort and rush to get stuff back into
370 product pipeline helped account for picking up bits & pieces of
release. recent thread about other pieces going out as resource manager
(another part of rushing things back into product pipeline, mvs/xa
effort convinced corp. to kill vm370, shutdown the vm370 development
effort, and move the people to POK to support mvs/xa ... in order to
make the mvs/xa ship schedule; endicott managed to save the product
mission ... but effectively had to reconstitute group from scratch).
http://www.garlic.com/~lynn/2009q.html#14 Electric Light Orchestra IBM song, in 1981?
a read-write subset (DWSS) was used later as part of original
relational/sql implementation ... system/r ... misc. past posts
mentioning system/r
http://www.garlic.com/~lynn/submain.html#systemr
and was supposed to have been part of tech. transfer of system/r to
endicott for SQL/DS ... but DWSS changes were dropped before SQL/DS
shipped and implementation had to be reworked to be done w/o it.
for random other trivia ... one of the people mentioned in this jan92
meeting ... claimed to have handled the tech transfer back to
STL for (mainframe) DB2
http://www.garlic.com/~lynn/95.html#13
some discussed in this recent post in comp.databases.theory
http://www.garlic.com/~lynn/2009p.html#82 What would be a truly relational operating system ?
http://www.garlic.com/~lynn/2009p.html#83 What would be a truly relational operating system ?
and of course
http://www.garlic.com/~lynn/2009p.html#43 From The Annals of Release No Sfotware Before Its Time
http://www.garlic.com/~lynn/2009p.html#46 From The Annals of Release No Sfotware Before Its Time
--
40+yrs virtualization experience (since Jan68), online at home since Mar1970
Anne & Lynn Wheeler
timothy...@US.IBM.COM (Timothy Sipples) writes:
> This cuts both ways. Obviously you're not going to support an OSA Express
> or FICON Express adapter on an X86 server. :-) But yes, if you're doing
> something like programming directly to an audio adapter in a PCI slot,
> that's not likely to be a System z-hosted program. (Though you never know
> as technology marches on.)
re:
http://www.garlic.com/~lynn/2009q.html#15 Mainframe running 1,500 Linux servers?
http://www.garlic.com/~lynn/2009q.html#19 Mainframe running 1,500 Linux servers?
recent posts/thread in comp.arch about early days of fiber-channel
standard and battles with mainframe channel engineers trying to layer
half-duplex protocol (FICON) on top of underlying full-duplex (actually
dual-simplex) infrastructure (i.e. half-duplex resulted in lower thruput
than running dual-simplex)
http://www.garlic.com/~lynn/2009p.html#84 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009p.html#85 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009q.html#0 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009q.html#1 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009q.html#2 Anyone going to Supercomputers '09 in Portland?
some of this goes back to days before even ESCON was released.
Anne & Lynn Wheeler
re:http://www.garlic.com/~lynn/2009q.html#20 Mainframe running 1,500 Linux servers?
basic ESCON technology had been knocking around POK for quite some
time before being released. One of the Austin engineers adopted it for
RS6000 ... making it about 10% faster (220mbits/sec instead of
200/mbits/sec) and actually full-duplex (so it had quite a bit higher
throughput) ... and used "Rochester" optic drivers ... which were quite a
bit cheaper;
http://www.garlic.com/~lynn/2009q.html#email870211
It made the SLA (serial-link-adaptors) for the RS6000 quite a bit
better than ESCON ... but incompatible. Anybody wanting to do ESCON
from RS6000 had to obtain it from completely different source.
In the late 80s, LANL was pushing standardization of the 100mbyte Cray
(half-duplex) channel as HIPPI and LLNL was pushing was serial
technology they had as 1gbit FCS. We were somewhat involved in both
activities ... examp
http://www.garlic.com/~lynn/2006x.html#email920129
The Austin SLA engineer had started looking at exhancing SLA to
800mbits/sec but we convinced him to switch to working on FCS instead.
There started to also be some ESCON participation ... looking to
overlay ESCON half-duplex on top of full-duplex FCS. This was somewhat
in line with work on "serial" HIPPI (i.e. moving Cray parallel
half-duplex "copper" 100mbyte/sec channel to fiber).
This old posts mentions jan92 meeting on doing 128-way RS6000
cluster scaleup
http://www.garlic.com/~lynn/95.html#13 SSA
Above also mentions moving Hursley 9333 disk subsystem (ran
encapsulated scsi protocols over full-duplex "serial" copper,
initially at 80mbits/sec ... full-duplex had several latency and
throughput advantages). This turned into IBM's "SSA" technology.
http://www-01.ibm.com/support/docview.wss?rs=505&uid=ssg1S1002348
SSA technology breaks 3000 IO bottlenecks
http://www.3000newswire.com/subscribers/SSAPrimer.html
We had suggested making it instead compatible with FCS protocol
... with an entry at 1/8th FCS over copper ... but allowing it to
scaleup to full FCS over fiber (and be able to interoperate in FCS
environment).
recent thread/post mentioning some of the above:
http://www.garlic.com/~lynn/2009q.html#0 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009q.html#1 Anyone going to Supercomputers '09 in Portland?
http://www.garlic.com/~lynn/2009q.html#2 Anyone going to Supercomputers '09 in Portland?
and from long ago and far away (think.com hosted standards mailing
list for both FCS and HIPPPI):
Date: 24 Oct 90 04:59:59 EDT
To: fc <fiber-ch...@Think.COM>
Subject: October Fiber Channel Minutes
TO: FC
October 23, 1990
TO: X3T9.3 Fiber Channel Working Group Members
FROM: Roger Cummings
SUBJECT: FIBER CHANNEL WORKING GROUP MINUTES
Please find attached a draft of the minutes of the ANSI X3T9.3
Fiber Channel Working Group of October 16 thru 18, 1990. Note
that there are also fifteen Attachments to the minutes that
relate to presentations at the meeting.
The next Fiber Channel Working Group meeting will be held on
November 1 and 2 as part of a Working Group week that is being
hosted by Bill Spence of Texas Instruments at the Stouffer Hotel
in Austin, TX. An announcement for this week is included. The
following working group meeting will be held on the Wednesday and
Thursday of the December plenary week (December 5 & 6) that is
being hosted by AMD at the Hyatt on First Street in San Jose, CA.
Note that the Working Group meeting will begin on Tuesday when
the X3T9.3 plenary finishes, and thus will be a three day
meeting. An announcement for the plenary week is attached.
A schedule of X3T9.3 meetings (both plenaries and working groups)
for the remainder of 1990 and the first half of 1991 is attached.
Note that it is also intended to hold a three day Fiber Channel
Working Group during the December plenary week. Hosts are still
required for the August, October and December plenary weeks in
1991 and the November 1991 Working Group week.
Note that ANSI has overruled as unconstitutional any attempt to
limit discussion on a subject to those who have brought
documentation on that subject.
It has been decided to maintain a standalone Document Register
for Fiber Channel, and the maintenance of such a register was
begun at the October meeting. The register will be divided by
subject and the numbering system used will be of the following
format:
FC-*/90-xxxRn
where:
* is the level to which it applies - 0, 1, 2, 3, 4, F or G
(If it applies across levels then it is a G for General)
xxx is the sequential number assigned by the Secretary
n is the revision level, beginning at base 0
Note that future presenters are strongly requested to have such a
document number and a page number on the top right-hand corner of
each page of their presentation (this requirement is waived for
the FC-0, -1 , -2, -3 and -4 documents themselves). They should
contact myself as indicated below to obtain the specific
sequential number.
Regards
1990 SCHEDULE OF MEETINGS
DATES EVENT LOCATION
10/29 - 11/02 Working Groups Austin, TX
12/03 - 12/06 Plenaries and Working Groups San Jose, CA
01/14 - 01/18 Working Groups Irvine, CA
02/18 - 02/20 Plenaries and Working Groups Austin, TX
03/18 - 03/22 Working Groups Cupertino, CA
04/22 - 04/25 Plenaries and Working Groups St. Petes, FL
05/20 - 05/24 Working Groups Wichita, KS
06/17 - 06/20 Plenaries and Working Groups Minneapolis, MN
The format of the week for months containing only working groups
is as follows:
Monday SCSI Common Access Method Group
Tuesday and Wednesday SCSI Working Group
HIPPI Working Group
Thursday and Friday Fiber Channel Working Group
The format of the week for the months containing both plenaries
and working groups is as follows:
Monday Separate X3T9.2 (SCSI) and
X3T9.3 plenaries (Monday
portion of X3T9.3 devoted to
IPI and HPPI).
Tuesday Separate X3T9.2 (SCSI) and
X3T9.3 plenaries (Tuesday
portion of X3T9.3 devoted to
Fiber Channel). Fiber Channel
Working Group begins when
X3T9.3 plenary ends (normally
before noon).
Wednesday Fiber Channel Working Group.
Thursday Fiber Channel Working Group
(December only).
DOCUMENT REGISTER
NUMBER TITLE
FC-0/90-001R0 FC-0 Revision 1.7
002R0 High performance LW Laser SM
003R0 Militarized Optoectronic Devi
FC-1/90-001R1.4 FC-1 Revision 1.4
FC-2/90-001R1.6 FC-2 Revision 1.6
002R0 Header Ideas and Proposal 10/
003R0 Proposed Header Changes 10/2
004R0 FC Header Proposal 10/14
005R0 Some Thoughts on Headers
006R0 Proposed Header
007R0 An R_RDY Primitive
008R0 Moving the Link/Device bit
009R0 Time Frames
FC-F/90-001R1 Fabric Requirements Revision
002R1 Event Log and Diagnostic Func
FC-G/90-001R0 Laser Focus World Article
002R0 Operation Level Proposal Outl
003R0 ESCON Architecture Slides
004R0 ESCON Converter and Director
Anne & Lynn Wheeler
edj...@PHOENIXSOFTWARE.COM (Edward Jaffe) writes:
> In his System z keynote address at SHARE in Austin, Karl Freund
> described a hypothetical future machine in which non-z processor
> "blades", mounted within the System z frame, would be available as
> "accelerators" to applications running on the traditional System z
> processors.
I have several papers that I wrote on the proposal from early 1985
... long-winded old post with reference:
http://www.garlic.com/~lynn/2004m.html#17 mainframe and microprocessor
other posts in this thread:
http://www.garlic.com/~lynn/2009q.html#15 Mainframe running 1,500 Linux servers?
http://www.garlic.com/~lynn/2009q.html#19 Mainframe running 1,500 Linux servers?
http://www.garlic.com/~lynn/2009q.html#20 Mainframe running 1,500 Linux servers?
http://www.garlic.com/~lynn/2009q.html#32 Mainframe running 1,500 Linux servers?