SAS Deserting the MF?

[Ed Gould]

SAS Makes Triple Play
SAS unleashed three major announcements at its Global Forum in San
Francisco this week, with a wave of new SAS High-Performance
Analytics capabilities, a newly unified SAS Customer Experience
marketing suite, and new levels of support for public and private
cloud deployment of SAS software.

SAS High-Performance Analytics software is designed to take advantage
of highly distributed, massively parallel processing (MPP) on memory-
intensive X86 servers. It has been a big strategic push for SAS over
the last two years, as customers demand ever-faster performance.

SAS previously offered High-Performance versions of industry specific
applications, such as financial risk analysis and marketing
optimization. But with the SAS High-Performance Analytics upgrades
announced this week and set for release in June, SAS will bring MPP
power to six core products in its portfolio: Statistics, Data Mining,
Text Mining, Optimization, Econometrics, and Forecasting.
READ THE ENTIRE ARTICLE AND POST YOUR THOUGHTS

http://www.informationweek.com/software/business-intelligence/sas-
makes-triple-play/240153951?pgno=2

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to list...@listserv.ua.edu with the message: INFO IBM-MAIN

[Charles Mills]

I don't read "SAS abandoning mainframe" -- I read "SAS emphasizing parallel
X86's for big data."

I'm not defending SAS, that's just how I read the story. Am I wrong?

Massively parallel X86's seems to be the vehicle of choice for big data
analysis:
http://www-01.ibm.com/software/data/db2/zos/analytics-accelerator/

Charles

[Bill Johnson]

If they are embracing x86 servers it seems to be a bad move. Isn't IBM trying to sell their x86 business? (to Lenovo) Mainly because the sales are dropping.


________________________________
From: Ed Gould <edgou...@COMCAST.NET>
To: IBM-...@LISTSERV.UA.EDU
Sent: Saturday, May 4, 2013 3:06 AM

Subject: SAS Deserting the MF?

[Anne & Lynn Wheeler]

mello...@YAHOO.COM (Bill Johnson) writes:
> If they are embracing x86 servers it seems to be a bad move. Isn't IBM
> trying to sell their x86 business? (to Lenovo) Mainly because the
> sales are dropping.

all hardware sales ... and x86 has the lowest profit margin, however

I.B.M., Missing Estimates, Falters as Slow Hardware Sales Hurt Revenue
and Profits
www.nytimes.com/2013/04/19/technology/ibm-shares-fall-after-earnings-miss-estimates.html

Lenovo, IBM Talks on Server Deal Said to Break Down
http://www.bloomberg.com/news/2013-05-02/lenovo-ibm-talks-on-server-deal-said-to-break-down.html

IBM was looking to divest a portion of its business with lower profit as
it seeks earnings per share of $20 by 2015, compared with $15.25 last
year. The company has had trouble selling its hardware products, with a
17 percent drop in the unit's revenue last quarter.

... snip ...

Stockman in "The Great Deformation: The Corruption of Capitalism in
America" ... talks about stock buybacks are a mini-form of LBO, with
the executives reaping huge rewards, pg457/loc9844-46:

The leader was ExxonMobil, which repurchased $160 billion of its own
shares during 2004-2011. It was followed by Microsoft at $100 billion,
IBM at $75 billion, and Hewlett-Packard, Proctor & Gamble, and Cisco
with $50 billion each. Even the floundering shipwreck of merger mania
known as Time Warner Inc. bought back $25 billion.

... snip ...

goes into lots of detail about executives managing their bonuses tied to
stock price via stock buybacks (reduces number of shares so
earnings/share goes up).

a little more from stockman, just a little on IBM last decade or so,
pg464/loc9995-10000:

IBM was not the born-again growth machine trumpeted by the mob of Wall
Street momo traders. It was actually a stock buyback contraption on
steroids. During the five years ending in fiscal 2011, the company spent
a staggering $67 billion repurchasing its own shares, a figure that was
equal to 100 percent of its net income.

pg465/10014-17:

Total shareholder distributions, including dividends, amounted to $82
billion, or 122 percent, of net income over this five-year
period. Likewise, during the last five years IBM spent less on capital
investment than its depreciation and amortization charges, and also
shrank its constant dollar spending for research and development by
nearly 2 percent annually.

... snip ...

my biggest quibble with stockman is that he glosses over the rating
agencies selling triple-A ratings on toxic CDOs (when they knew they
weren't worth triple-A, from congressional Oct2008 hearings). Those
triple-A ratings significantly enabled the over $27T done during the
bubble ... and that $27T significantly dwarfs many of the other issues
he cites. reference to over $27T
http://www.bloomberg.com/apps/news?pid=20601109&refer=home&sid=a0jln3.CSS6c

issue is that recently x86 makers have said that they are shipping more
x86 server chips to the cloud operators ... than are going to brand name
server vendors (Dell, HP, IBM, etc) ... it isn't that x86 servers are
declining ... the numbers are exploding ... but it is declining market
for the brand name vendors. some recent posts mentioning z196 compared
to e5-2600 blade (x86 server, staple of large cloud operators).
http://www.garlic.com/~lynn/2013.html#10 From build to buy: American Airlines changes modernization course midflight
http://www.garlic.com/~lynn/2013.html#16 From build to buy: American Airlines changes modernization course midflight
http://www.garlic.com/~lynn/2013.html#17 Still think the mainframe is going away soon: Think again. IBM mainframe computer sales are 4% of IBM's revenue; with software, services, and storage it's 25%
http://www.garlic.com/~lynn/2013b.html#5 mainframe "selling" points
http://www.garlic.com/~lynn/2013b.html#6 mainframe "selling" points
http://www.garlic.com/~lynn/2013b.html#7 mainframe "selling" points
http://www.garlic.com/~lynn/2013b.html#8 mainframe "selling" points
http://www.garlic.com/~lynn/2013b.html#10 FW: mainframe "selling" points -- Start up Costs
http://www.garlic.com/~lynn/2013b.html#15 A Private life?
http://www.garlic.com/~lynn/2013c.html#59 Why Intel can't retire X86
http://www.garlic.com/~lynn/2013c.html#60 Why Intel can't retire X86
http://www.garlic.com/~lynn/2013c.html#63 What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013c.html#68 relative mainframe speeds, was What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013c.html#84 What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013d.html#5 What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013d.html#16 relative mainframe speeds, was What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013e.html#4 Oracle To IBM: Your 'Customers Are Being Wildly Overcharged'
http://www.garlic.com/~lynn/2013f.html#35 Reports: IBM may sell x86 server business to Lenovo
http://www.garlic.com/~lynn/2013f.html#37 Where Does the Cloud Cover the Mainframe?
http://www.garlic.com/~lynn/2013f.html#38 Reports: IBM may sell x86 server business to Lenovo
http://www.garlic.com/~lynn/2013f.html#51 Reports: IBM may sell x86 server business to Lenovo
http://www.garlic.com/~lynn/2013f.html#57 The cloud is killing traditional hardware and software
http://www.garlic.com/~lynn/2013f.html#64 What Makes an Architecture Bizarre?
http://www.garlic.com/~lynn/2013f.html#70 How internet can evolve
http://www.garlic.com/~lynn/2013f.html#72 The cloud is killing traditional hardware and software
http://www.garlic.com/~lynn/2013f.html#73 The cloud is killing traditional hardware and software
http://www.garlic.com/~lynn/2013f.html#74 The cloud is killing traditional hardware and software
http://www.garlic.com/~lynn/2013g.html#2 A Complete History Of Mainframe Computing

--
virtualization experience starting Jan1968, online at home since Mar1970

[R.S.]

W dniu 2013-05-05 03:14, Bill Johnson pisze: > If they are embracing x86 servers it seems to be a bad move. Isn't > IBM trying to sell their x86 business? (to Lenovo) Mainly because the > sales are dropping. No. x86 market (or Intel market or just PC market to honour AMD and x64) consist of personal computers (desktops, laptops) and servers. While personal computers are partially being replaced with some mobile devices, the PC servers take marketshare from HP, Sun, p Series and (small) mainframes. Side notes: - Having significant marketshare doesn not mean having profits. - Server market may shrink, despite of the platform. Or at least not to grow so rapid. (Did anyone said recession?) - IBM sold personal PCs, not servers. - (last but not least) IBM also sold personal printers (Lexmark), big printers (Ricoh), ATMs (Diebold?), networking devices (Cisco), hard disks (Hitachi), printed circuits (Celestica), etc. IMHO neither the buyers want to have loss from inprofitable business, nor the business is vanishing. We still use ATMs, hard drives, printers. Radoslaw Skorupka Lodz, Poland tej wiadomo ci mo e zawiera informacje prawnie chronione Banku przeznaczone wy cznie do u ytku s bowego adresata. Odbiorc e by jedynie jej adresat z wy czeniem dost pu os b trzecich. Je eli nie jeste adresatem niniejszej wiadomo ci lub pracownikiem upowa nionym do jej przekazania adresatowi, informujemy, e jej rozpowszechnianie, kopiowanie, rozprowadzanie lub inne dzia anie o podobnym charakterze jest prawnie zabronione i mo e by karalne. Je eli otrzyma wiadomo omy kowo, prosimy niezw ocznie zawiadomi nadawc wysy c odpowied oraz trwale usun wiadomo czaj c w to wszelkie jej kopie wydrukowane lub zapisane na dysku. This e-mail may contain legally privileged information of the Bank and is intended solely for business use of the addressee. This e-mail may only be received by the addressee and may not be disclosed to any third parties. If you are not the intended addressee of this e-mail or the employee authorised to forward it to the addressee, be advised that any dissemination, copying, distribution or any other similar activity is legally prohibited and may be punishable. If you received this e-mail by mistake please advise the sender immediately by using the reply facility in your e-mail software and delete permanently this e-mail including any copies of it either printed or saved to hard drive. BRE Bank SA, 00-950 Warszawa, ul. Senatorska 18, tel. +48 (22) 829 00 00, fax +48 (22) 829 00 33, www.brebank.pl, e-mail: in...@brebank.pl d Rejonowy dla m. st. Warszawy XII Wydzia Gospodarczy Krajowego Rejestru S dowego, nr rejestru przedsi biorc w KRS 0000025237, NIP: 526-021-50-88. ug stanu na dzie 01.01.2013 r. kapita zak adowy BRE Banku SA (w ca ci wp acony) wynosi 168.555.904 z otych. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to list...@listserv.ua.edu with the message: INFO IBM-MAIN

[Paul Gilmartin]

On Sat, 4 May 2013 21:53:08 -0400, Anne & Lynn Wheeler wrote:
> ...

> http://www.garlic.com/~lynn/2013f.html#57 The cloud is killing traditional hardware and software

> ...
>
"... is killing traditional ..." is a paraphrase of "progress". In the 1950s
you might have heard "The electronic computer is killing traditional
punched card tabulators." Or in the 1960s "The transistor is killing
traditional vacuum tube computers."

This is unpleasant only to those overly invested in the obsolescent
technologies. And not all such technological prognostications become
reality. Magnetic bubbles. Cryogenic computers. Quantum computers.

-- gil

[Anne & Lynn Wheeler]

PaulGB...@AIM.COM (Paul Gilmartin) writes:
> "... is killing traditional ..." is a paraphrase of "progress". In the 1950s
> you might have heard "The electronic computer is killing traditional
> punched card tabulators." Or in the 1960s "The transistor is killing
> traditional vacuum tube computers."
>
> This is unpleasant only to those overly invested in the obsolescent
> technologies. And not all such technological prognostications become
> reality. Magnetic bubbles. Cryogenic computers. Quantum computers.

re:
http://www.garlic.com/~lynn/2013g.html#5 SAS Deserting the MF?

part of the issue is that the x86 server chip makers are saying they
ship more x86 server chips to cloud operators than they are shipping to
the brand name server vendors ... and those cloud volumes aren't
included in the x86 server market numbers (aka cloud x86 servers are
larger than the total of brand name x86 server "market").

another part of the issue is that the cloud operators for a decade or so
have been claiming they build their own servers for 1/3rd the price of
the same servers from brand name x86 server vendors (contributed to huge
downward pressure on profit margins in the rest of the market). There is
even rumor that some of the brand name vendors have gotten into this
extremely low margin business ... assembling x86 components for cloud
operations.

when something becomes larger than what is thought to be the major
market ... then it may be time to pay some attention.

ibm has $1815 base price for e5-2600 blade (a common x86 server found at
cloud operator). it is two 8-core chips for 16 processors and benchmark
of 527BIPS or $3.44/BIPS (compared to 80processor z196 at 50BIPS @$28M
or $560,000/BIPS); cloud vendors claims of assembling servers at 1/3rd
the price of brand name vendors would possibly bring it close to
$1/BIPS.

e5-2600v2 with new chip technology due out this year is predicted to be
twice the performance and 12cores/chip ... bringing e5-2600 blade to
over 1TIPS. There is also references to e5-4600 blade in same form
factor with four chips ... which could be well over 2TIPS.

aggregate mainframe sales for the past several years (before the latest
decline) has been approx. equivalent to 180 80-processor z196 and
@50BIPS comes out to be about 9TIPS/year ... which is less than half a
rack of e5-2600 blades.

big cloud operators are building new megadatacenters with hundreds of
thousands of such blades (and millions of processors).

last year financials had mainframe processors 4% of revenue but total
mainframe business (with software, services, etc) was 25% of revenue
(and 40% of profit). that has mainframe customers paying ibm avg. of
6.25 times their processor purchase for mainframe operations ... i.e.
TCO just to IBM for $28M 80processor, 50BIP z196 then averages total of
$175M or $3.5M/BIPS (compared to possibly $1/BIPS at cloud operations, a
factor of 3,500,000 times difference).

--
virtualization experience starting Jan1968, online at home since Mar1970

[Timothy Sipples]

I must take issue with the "BIPS" measurement and the cross-architecture
comparisons presented in this discussion. They're extremely misleading at
best. "MIPS" and "BIPS" are perilous enough within zEnterprise capacity
estimations, but they go haywire rapidly when carried elsewhere.

That should be obvious from the suggestion that half a rack of blade
servers is in any way comparable to 180 z196 machines with 80 cores each.
In a word, no.

An 80 core z196 doesn't have only 80 cores, by the way.

--------------------------------------------------------------------------------------------------------
Timothy Sipples
GMU VCT Architect Executive (Based in Singapore)
E-Mail: sip...@sg.ibm.com

[Elardus Engelbrecht]

Timothy Sipples wrote:

>I must take issue with the "BIPS" measurement and the cross-architecture comparisons presented in this discussion. They're extremely misleading at best. "MIPS" and "BIPS" are perilous enough within zEnterprise capacity estimations, but they go haywire rapidly when carried elsewhere.

Hmmm, I note your comment, but I want to give you some TIPS, what about TIPS? (Terabyte Instructions per Seconds)

Is that perilous enough? ;-D

Back to square one where k, kibi and friends reside... ;-D

<shields and phasers up while running back to hide under my rock>

;-)

Groete / Greetings
Elardus Engelbrecht

[Anne & Lynn Wheeler]

sip...@SG.IBM.COM (Timothy Sipples) writes:
> I must take issue with the "BIPS" measurement and the cross-architecture
> comparisons presented in this discussion. They're extremely misleading at
> best. "MIPS" and "BIPS" are perilous enough within zEnterprise capacity
> estimations, but they go haywire rapidly when carried elsewhere.

re:
http://www.garlic.com/~lynn/2013g.html#5 SAS Deserting the MF?

http://www.garlic.com/~lynn/2013g.html#7 SAS Deserting the MF?

industry standard benchmark dhrystone used for MIPS, BIPS, TIPS
... which is not actual instructions/sec ... but number of
iterations/sec scaled to 370/158-3 assumed to be 1MIPS (aka ibm
mainframe is used as industry standard baseline for MIPS, BIPS, & TIPS)
http://en.wikipedia.org/wiki/Instructions_per_second

there is also FLOPS ... floating point operations/sec
http://en.wikipedia.org/wiki/FLOPS

other standard benchmarks are SPEC
http://www.spec.org/benchmarks.html

and TPC.
http://www.tpc.org/

In many cases, IBM has hundreds of published, certified industry
standard benchmarks ... just none for recent mainframes. recent post
discussing subject
http://www.garlic.com/~lynn/2013b.html#25 Still think the mainframe is going away soon: Think again. IBM mainframe computer sales are 4% of IBM's revenue; with software, services, and storage it's 25%

other recent posts discussing subject:

http://www.garlic.com/~lynn/2013b.html#6 mainframe "selling" points

http://www.garlic.com/~lynn/2013c.html#84 What Makes an Architecture Bizarre?

http://www.garlic.com/~lynn/2013c.html#88 What Makes an Architecture Bizarre?

http://www.garlic.com/~lynn/2013d.html#5 What Makes an Architecture Bizarre?

http://www.garlic.com/~lynn/2013f.html#80 The cloud is killing traditional hardware and software

basis for comparison, cite any industry standard benchmark for current
ibm mainframe.

--
virtualization experience starting Jan1968, online at home since Mar1970

[Anne & Lynn Wheeler]

re:
http://www.garlic.com/~lynn/2013g.html#5 SAS Deserting the MF?
http://www.garlic.com/~lynn/2013g.html#7 SAS Deserting the MF?

http://www.garlic.com/~lynn/2013g.html#9 SAS Deserting the MF?

one of the claims about x86 performance increase in BIPS over the past
decade has been attributed to competition between multiple vendors
producing x86 chips. the other claim is that for the past several x86
chip generations they've gone to RISC cores with hardware layer that
translate x86 instructions into RISC micro-ops (for decades, RISC
performance has been significantly better than x86, but x86 move to
RISC core appears to be negating that difference).

base-line of dhrystone is

370/158-3 1 processor, 1MIPS, (1MIPS/proc), 1972

other numbers approx

370/168-3 1 processor 3MIPS, (3MIPS/proc),
3033 1 processor 4.5MIPS, (4.5MIPS/proc),
3081D 2 processor 10MIPS, (5MIPS/proc),
3081K 2 processor 14MIPS, (7MIPS/proc),

this is recent post discussing reasons for 3081 throughput being much
slower than claimed (& 3090 being first "real" new IBM 370 after 168)
http://www.garlic.com/~lynn/2013g.html#1
http://www.garlic.com/~lynn/2013g.html#8

This source has MIPS vendor claims
http://www.roylongbottom.org.uk/mips.htm

3090-600E 6 processor, 89MIPS (17MIPS/proc)
ES/9000-982 8 processor, 408MIPS (51MIPS/proc)
9672 G6 ZZ7 12 processor, 1644MIPS (137MIPS/proc)

the numbers appear to show increase consistent through the generations
from the 370/158-3 base ... which is also used as baseline for the
industry standard dhrystones benchmark.

older post here in ibm-main looking at last decade or so, from ibm
publications & website
http://www.garlic.com/~lynn/2013b.html#45

z900, 16 processors, 2.5BIPS (156MIPS/proc), Dec2000
z990, 32 processors, 9BIPS, (281MIPS/proc), 2003
z9, 54 processors, 18BIPS (333MIPS/proc), July2005
z10, 64 processors, 30BIPS (469MIPS/proc), Feb2008
z196, 80 processors, 50BIPS (625MIPS/proc), Jul2010
EC12, 101 processors, 75BIPS (743MIPS/proc), Aug2012

this post
http://www.garlic.com/~lynn/2013g.html#1

also discusses killing of acs-360 effort in late 60s because company was
worried that it would advance computing technology too fast, threatening
the company's control of the market.
http://people.cs.clemson.edu/~mark/acs_end.html

the bottom of the above page discusses acs-360 (from late 60s)
features finally showing up in ES/9000 in the early 90s.

--
virtualization experience starting Jan1968, online at home since Mar1970

[J. Leslie Turriff]

On 2013-05-06 05:42:47 Elardus Engelbrecht wrote:
> Hmmm, I note your comment, but I want to give you some TIPS, what about
> TIPS? (Terabyte Instructions per Seconds)

I think you mean Trillions of Instructions Per Second. :-)

Leslie