Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

[Joseph G. Baron]

Two interesting points from Intel’s Jason Waxman on private cloud and cost. He cautions that building a private cloud is “quite a daunting task”, that there is no off-the-shelf hardware and interconnect solution suitable for a private cloud. He also offers a view on conventional IT costs: servers are 50% of data center outlays, electrical power and cooling are another 23%.

 

This suggests to me a baseline comparison of traditional IT vs. (public) cloud costs: if a unit of incremental computing capacity – say a single server – costs X per hour, then the equivalent capacity in the cloud should cost at least 2X per hour. Then the other values of cloud (non-ownership, elasticity etc.) should be added *on top* of the base cloud cost.

 

The article is “Intel Details Private Cloud Hurdles”, Information Week reporting on the Cloud Computing Conference & Expo in Santa Clara.

 

(Maybe some folks on the list are there in person, but I have to be content with the trade press reports.)

 

Regards,

Joe

_____________________________________________

Joseph G. Baron, Enterprise IT Solutions Architect

11808 Mountbatten Way, Raleigh, NC 27613

Mobile / Work: +1 (919) 809-9542

jgb...@gmail.com    http://www.linkedin.com/in/jgbaron

 

[Ray DePena]

Hi Joe,

Only 27% split between labor, storage, networking, and other costs?

"He also offers a view on conventional IT costs: servers are 50% of data center outlays, electrical power and cooling are another 23%."

--
Ray DePeña
Director, Stealth Startups
Strategic Business Advisor

http://www.linkedin.com/in/raydepena
Sacramento, CA 95630
(916) 941-5558

[Rao Dronamraju]

Interesting points from Jason Glassman!. But I think he and Intel have two options.

 

1)       Figueout a way to make the daunting tasks un-daunting in building private clouds or

2)       Hire a bunch of Psychologists to over come the CIO FUD! With public clouds.

 

Since Intel is almost a monopoly and has lot of money, it probably can do bothJ

 

---

[Ray Nugent]

Maybe in TX. Most everywhere else P&C is closer to 40-45%

Ray

---

From: Ray DePena <ray.d...@gmail.com>
To: cloud-c...@googlegroups.com
Sent: Wed, November 4, 2009 4:55:49 PM
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

[Jan Klincewicz]

So does Intel have a vested interest im maintaining the status quo (server sprawl) ? Would CC (public) favor less expensive AMD procs?

---

From: Rao Dronamraju <rao.dro...@sbcglobal.net>
Sent: Wednesday, November 04, 2009 8:10 PM
To: cloud-c...@googlegroups.com

Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

Interesting points from Jason Glassman= !. But I think he and Intel have two options.

 

1)      = Figueout a way to make the daunting tasks un-daunting in buildi= ng private clouds or

2)      = Hire a bunch of Psychologists to over come the CIO FUD! With pu= blic clouds.

 

Since Intel is almost a monopoly and h= as lot of money, it probably can do bothJ

 

---

From:cloud-c...@googlegroups.com [mailto:cloud-c...@googlegroups.com] = On Behalf Of Joseph G. Baron

Sent: Wednesday, November 04= , 2009 4:08 PM
To: cloud-c...@googlegroups.com

Subject: [ Cloud Computing ] Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data=

 

Two interesting points from Intel’s Jason Waxman on private cloud and cos= t. He cautions that building a private cloud is “quite a daunting task”, that there is no off-the-shelf hardware and interconnect solut= ion suitable for a private cloud. He also offers a view on conventional IT cost= s: servers are 50% of data center outlays, electrical power and cooling are another 23%.

 

This suggests to me a baseline comparison of traditional IT vs. (public) cloud costs: if a unit of incremental computing capacity – say a single ser= ver – costs X per hour, then the equivalent capacity in the cloud should = cost at least 2X per hour. Then the other values of cloud (non-ownership, elasti= city etc.) should be added *on top* of the base cloud cost.

 

The article is “Intel Details Private Cloud Hurdles”, Information Week reporting on the Cloud Computing Conference & Expo in Santa Clara.

 

(Maybe some folks on the list are there in person, but I have to be content with t= he trade press reports.)

[Ray]

Of course. Storage and interconnect will emerge as the critcal areas in cloud infrustructure while CPU fades in importance for the first phases.

Ray

[Hilly Biker]

@Joe

Correction, Intel made it clear that these numbers were specifically
for Internet Data Centers (ie. ones packed to the gills with fairly
dense (1U, 2U and blades) equipment ... which is considerably
different and distinct from traditional data centers.

There is good data from The Green Grid, Uptime Institute, 451,
Forrester, etc about the average cost breakdowns for a traditional
enterprise DC.

However, Intel used data for IDCs because they are closest to what a
cloud DC would look like, so if you are looking at putting together a
private, internal cloud your DC may look more like a traditional one
now, but after a couple of technical refreshes (off lease, updating
equipment, etc) it will lean more and more towards the IDC model (up
to the limits of your DC's power and cooling delivery capabilities, of
course).

-Phil

On Wed, Nov 4, 2009 at 2:08 PM, Joseph G. Baron <jgb...@gmail.com> wrote:

[Ray DePena]

What are we saying here?  These numbers just don't seem to add up.  Perhaps labor costs are not included at all which is not that uncommon to see as companies often times don't account for those costs (I've been party to several cases where labor costs were not accounted for).

If,

P&C is 40-45% and
Servers are 50%

Am I to believe that labor, storage, networking and other associated costs to run a datacenter are a mere 5%?

None of those numbers line up with any of the data center and network outsourcing experiences I've had.  Nor for any of the services we offered.  Systems (servers and storage) typically amounted to 15-25% of the deals I saw.  Fully burdened labor headcount was never a trivial amount, but I'm to believe these numbers, they're less than 5%?

[Joseph G. Baron]

On Nov 4, 2009, at 7:55 PM, Ray DePena <ray.d...@gmail.com> wrote:

> Only 27% split between labor, storage, networking, and other costs?

Ray,

It's not clear whether or not labor was included; I suspect not. The
article is pretty sketchy, maybe somebody who actually attended the
talk could fill in the details...

— Joe
Sent from my iPhone (919) 809-9542

[Joseph G. Baron]

On Nov 4, 2009, at 8:21 PM, Ray Nugent <rnu...@yahoo.com> wrote:

> Maybe in TX. Most everywhere else P&C is closer to 40-45%

Ray,

That 23% seemed low to me as well, I'm used to seeing numbers that put
P&C closer to your 40-45% -- that is, about the same as (or even more
than) the cost of the server. But 23% is Waxman's number.

[Rao Dronamraju]

Possibly….but AMD is too small to make a dent on Intel at this time.

AMD needs a major M&A partner….nobody is bold enough to acquire AMD and pissoff Intel…

Microsoft atleast has some competition from Java/IBM/Sun-Oracle camp but Intel virtually has no competition.

 

---

[Ray DePena]

Perhaps Intel with its 80%+ share has gotten to the point where Ma Bell was years ago before the breakup..... 

Do you think if Intel were broken up that there would be more business?

[Miha Ahronovitz]

Phil,

Not sure what IDC data you are referring to. The latest report from IDC HPC
Market update is about 3 weeks old

IDC market update for HPC shows the cost per core $0.8K per core. Compared
to the cost of $2.4K per CPU, IDC assumed an average of 3 core per CPU in
HPC installations sold. This includes all Hw, IB or other HPC must-have
add-on. Plus system software. Not labor or cooling or utilities

Assuming these cost a 30,000 cores (10,000 CPU) sells for 24M. Assuming a
lease of 3 years, at 0.085 per hour per node instance (which is one core) ,
at 100% utilization the gross profit margin is 94%. For a five year lease,
the gross profit margin jumps to 98%.

The cloud computing model is so compelling, that one wonders why we still
buy hardware as a classical DC, and not offer directly services via clouds?
Why people buy computers? To hang them on the wall like masterpiece pictures
or place them in garages, like a collection of automobiles? :-) No.
Companies need to offer application to a constituency of users and cloud
computing model is so compelling.

Now we have to decide whether we want to let Amazon collect all money on the
table, or it is better to re-organize the IT in private clouds. The case for
private clouds is overwhelming. Sure, the only drawback is that a private
cloud has limited elasticity. When demand is very high, we need a hybrid
cloud to add AWS external resources for peaks of short duration. Short
duration must be defined. If the bill that the hybrid cloud owners pay
Amazon is very high, one can calculate when the private cloud must enlarged.

I don't know whether it makes a difference whether you have an Internet Data
Center of an HPC Data center. I think not. The conclusions are the same, IMO

Miha

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Hilly Biker
Sent: Wednesday, November 04, 2009 6:50 PM
To: cloud-c...@googlegroups.com

[Alan Ho]

My 2 cents with regards to the private clouds is that the biggest cost
savings will come through standard virtualization - which the
technology is basically free today with xen and open sourcing of
VMware's ESX product. The second biggest win is putting in management
software like VMware vSphere for managing traditional DC - that too
might become free as VMware moves up the stack yet again.

Regards,
Alan Ho

On Wednesday, November 4, 2009, Rao Dronamraju

<rao.dro...@sbcglobal.net> wrote:
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> Possibly….but AMD is too small to
> make a dent on Intel at this time.
>
> AMD needs a major M&A partner….nobody
> is bold enough to acquire AMD and pissoff Intel…
>
> Microsoft atleast has some competition
> from Java/IBM/Sun-Oracle camp but Intel virtually has no competition.
>
>
>
>
>
>
>
>
>
>
>

> From: cloud-c...@googlegroups.com [mailto:cloud-c...@googlegroups.com] On Behalf Of Jan Klincewicz
> Sent: Wednesday, November 04, 2009
> 8:03 PM
> To: cloud-c...@googlegroups.com
> Subject: [ Cloud Computing ] Re:
> Intel's Waxman: "[building a private cloud is] quite a daunting
> task", plus some cost data
>
>
>
>
>
> So does Intel have a vested interest
> im maintaining the status quo (server sprawl) ? Would CC (public) favor less
> expensive AMD procs?
>
>
>
>
>
>
>
>
>

> From: Rao
> Dronamraju <rao.dro...@sbcglobal.net>
> Sent: Wednesday,
> November 04, 2009 8:10 PM
> To: cloud-c...@googlegroups.com
> Subject: [ Cloud Computing
> ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting
> task", plus some cost data
>
>
>
> Interesting points from Jason
> Glassman= !. But I think he and Intel have two options.
>
>
>
> 1)      = Figueout
> a way to make the daunting tasks un-daunting in buildi= ng private clouds or
>
> 2)      = Hire a
> bunch of Psychologists to over come the CIO FUD! With pu= blic clouds.
>
>
>
> Since Intel is almost a monopoly
> and h= as lot of money, it probably can do bothJ
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

[Jan Klincewicz]

Intel seems to consider itself much more than a chipmaker.  Granted, they do a lot of R&D and platform stuff, etc. but at the end of the day, silicon pays the bills. Corporate Data Centers are more likely to care about the "Intel Inside" branding than CPs where the customer does not know or care what CPU is in the server.

Until Nehalem, Intel was at a significant disadvantage to AMD in the MIPS per WATT arena.  Dense DCs choosing blades frequently chose Opterons solely because of the thermal envelope.

I guess the only lesson I am learning is  not to trust "statistics" and proclamations about Cloud made by authorities who obviously have a few cards in the game.  That includes most of the current generation of IT vendors.

--
Cheers,
Jan

[Jan Klincewicz]

Alan:

Can you point me to any info regarding ESX being OPen Source ?  I had read articles a year o so ago about Maritz toying with the idea.  In my recent encounters with VMWare, they appeared adamantly opposed to the idea of "hypervisor as a commodity" as espoused by Citrix and Microsoft.  They still don;t seem to be hurting in market share, despite Citrix and MSFT enhancing their "free" products and that is probably in part due to the robustness of their management stack and ecosystem.

What would they have to gain by giving away what they currently sell at huge margin ?

P.S. I agree with you that the cost savings found in virtualizing  (at least at this juncture in time) will be the most significant benefit to "private" clouds.

<My 2 cents with regards to the private clouds is that the biggest cost
savings will come through standard virtualization - which the
technology is basically free today with xen and open sourcing of
VMware's ESX product. The second biggest win is putting in management
software like VMware vSphere for managing traditional DC - that too
might become free as VMware moves up the stack yet again. >>

--
Cheers,
Jan

[Hilly Biker]

@Miha,

Sorry for the confusion, I spelled it out once (Internet Data Center)
but didn't indicate that I would be using IDC as the initials for it
afterwards, which means that it was easily confused with the
marketing/consulting outfit. In general, this means the highly
specialized data centers such as Google or Amazon.

That said, I agree with the rest of your post ... Private cloud is
coming on strong due to many factors, some of which you refer to as
well as others on these various threads. And of course one of the most
powerful forces is the applications themselves. There are still many
shops out there that take 12-18 months to move to a new version of an
OS or sometimes even major new patches because of all of the
background tst/qa that has to be done.

In fact, most of the data that I've seen says that companies have,
roughly, about an average of the same number of virtual instances
compared to physical systems running applicaitons. That means that
there is still a huge amount of legacy code that will still take quite
a bit of work to get into any form of cloud, but the business case is
generally compelling enough that most companies will/are make the
effort.

-Phil

[Rao Dronamraju]

“Do you think if Intel were broken up that there would be more business?”

 

Not sure. If Intel is broken up lile AT&T, this would give the proprietary HW vendors (IBM,HP&Sun) a second chance.

Is it a good thing?...Don’t know, unless the proprietary HW prices come down to the level of commodity HW.

So far Intel is keeping the prices low despite being an almost monopoly.

If the proprietory HW vendors close their (HW) shops and Intel raises/controls the prices, then the Fed anti-trust might kick-in and Intel will might be either broken up or price regulated especially if computing becomes utilities.

 

 

---

[Rao Dronamraju]

“Until Nehalem, Intel was at a significant disadvantage to AMD in the MIPS per WATT arena.  Dense DCs choosing blades frequently chose Opterons solely because of the thermal envelope.”

 

Sure, technologically AMD had done a great job especially never letting Intel become complacent, but it is unable to steal the market share from Intel.

There are other things that might come into play when you make Cloud scale business decisions. How long will AMD last?....if you buy 50 to 100,000+ servers from AMD for your cloud and AMD files for bankruptcy, isn’t your cloud in a mega melt down?....

[Ray Nugent]

No, I did not include labor and in P&C I include the P&C infrastructure. Here's one man's breakdown -

http://perspectives.mvdirona.com/2008/11/28/CostOfPowerInLargeScaleDataCenters.aspx

Ray

---

From: Ray DePena <ray.d...@gmail.com>
To: cloud-c...@googlegroups.com

Sent: Wed, November 4, 2009 7:25:04 PM

[Jim Starkey]

Rao Dronamraju wrote:
>
> Possibly….but AMD is too small to make a dent on Intel at this time.
>
> AMD needs a major M&A partner….nobody is bold enough to acquire AMD
> and pissoff Intel…
>
> Microsoft atleast has some competition from Java/IBM/Sun-Oracle camp
> but Intel virtually has no competition.
>

Uh, do remember AMD64 vs. the Itanium (aka Itanic). AMD gave Intel
enough competition that Intel had to play dirty (and got thenselves into
a passle of world wide anti-trust trouble) until they could get their
technology back on track. At the moment, Intel seems to be ahead on
technology.

Intel vs. AMD seems like a pretty pragmatic market. Sun, for example,
sold AMD in addition to Sparc until Intel caught up, now sells all three.

And lets give a little credit to Microsoft who held the shotgun on Intel
until they agreed to adopt the AMD64 instruction set. As long as AMD and
Intel are software compatible (and Intel stops playing dirty), we've got
competition, and we're all benefiting.
>
> ------------------------------------------------------------------------
>
> *From:* cloud-c...@googlegroups.com
> [mailto:cloud-c...@googlegroups.com] *On Behalf Of *Jan Klincewicz
> *Sent:* Wednesday, November 04, 2009 8:03 PM
> *To:* cloud-c...@googlegroups.com
> *Subject:* [ Cloud Computing ] Re: Intel's Waxman: "[building a

> private cloud is] quite a daunting task", plus some cost data
>
> So does Intel have a vested interest im maintaining the status quo
> (server sprawl) ? Would CC (public) favor less expensive AMD procs?
>

> ------------------------------------------------------------------------
>
> *From: *Rao Dronamraju <rao.dro...@sbcglobal.net>
> *Sent: *Wednesday, November 04, 2009 8:10 PM
> *To: *cloud-c...@googlegroups.com
> *Subject: *[ Cloud Computing ] Re: Intel's Waxman: "[building a

> private cloud is] quite a daunting task", plus some cost data
>
> Interesting points from Jason Glassman= !. But I think he and Intel
> have two options.
>
> 1) = Figueout a way to make the daunting tasks un-daunting in buildi=
> ng private clouds or
>
> 2) = Hire a bunch of Psychologists to over come the CIO FUD! With pu=
> blic clouds.
>
> Since Intel is almost a monopoly and h= as lot of money, it probably
> can do bothJ
>

> ------------------------------------------------------------------------
>
> *From:*cloud-c...@googlegroups.com
> [mailto:cloud-c...@googlegroups.com] = *On Behalf Of *Joseph G. Baron
> *Sent:* Wednesday, November 04= , 2009 4:08 PM
> *To:* cloud-c...@googlegroups.com
> *Subject:* [ Cloud Computing ] Intel's Waxman: "[building a private

> cloud is] quite a daunting task", plus some cost data=
>
> Two interesting points from Intel’s Jason Waxman on private cloud and
> cos= t. He cautions that building a private cloud is “quite a daunting
> task”, that there is no off-the-shelf hardware and interconnect solut=
> ion suitable for a private cloud. He also offers a view on
> conventional IT cost= s: servers are 50% of data center outlays,
> electrical power and cooling are another 23%.
>
> This suggests to me a baseline comparison of traditional IT vs.
> (public) cloud costs: if a unit of incremental computing capacity –
> say a single ser= ver – costs X per hour, then the equivalent capacity
> in the cloud should = cost at least 2X per hour. Then the other values

> of cloud (non-ownership, elasti= city etc.) should be added **on top**

> of the base cloud cost.
>

> The article is “Intel Details Private Cloud Hurdles”, <3D> Information

> Week reporting on the Cloud Computing Conference & Expo in Santa Clara.
>
> (Maybe some folks on the list are there in person, but I have to be
> content with t= he trade press reports.)
>
> Regards,
>
> Joe
>
> _____________________________________________
>
> Joseph G. Baron, Enterprise IT Solutions Architect
>
> 11808 Mountbatten Way, Raleigh, NC 27613
>
> Mobile / Work: +1 (919) 809-9542
>
> jgb...@gmail.com http://www.linkedin.com/in/jgbaron
>
>
>
>
>
>
>
> >
>

--
Jim Starkey
Founder, NimbusDB, Inc.
978 526-1376

[Ray Nugent]

Free. Which is different than Open Source - http://www.vmware.com/products/esxi/

Ray

---

From: Jan Klincewicz <jan.kli...@gmail.com>
To: cloud-c...@googlegroups.com
Sent: Thu, November 5, 2009 5:54:28 AM

[Jan Klincewicz]

@Jim

I saw first-hand what Intel could do when they were first threatened by Opteron.  I forget all the stages, but disbelief and anger were there.  Everything but acceptance.  They should be thankful (as MSFT  is thankful of MAC) for providing hust enough competition to keep the DOJ off their butts.  Europe is all over them.

Also,I was competing internally with the margin demons in "Business Critical Servers" who were badmouthing Opteron's "64-bitness" compared to Itanic to my customers !!

I don't know how they are at all comparable to a public utility like AT&T (which WAS broken up.)  I think they are more like IBM was (where the Justice department was eventually outgunned.)

I really wish the best for AMD (and I respect Intel as well, but they were real babies when they got blindsided.)    There are some schisms emerging, especially in the Virtualization space where there is different page table handling etc.   This means when you pick a hypervisor, and build a farm, you pick a platform.  You won't be migrating live VMs from Opteron to Xeon without holding your breath and praying...


***************************************************************************************************************************************************************************************

Uh, do remember AMD64 vs. the Itanium (aka Itanic). AMD gave Intel
enough competition that Intel had to play dirty (and got thenselves into
a passle of world wide anti-trust trouble) until they could get their
technology back on track. At the moment, Intel seems to be ahead on
technology.

Intel vs. AMD seems like a pretty pragmatic market. Sun, for example,
sold AMD in addition to Sparc until Intel caught up, now sells all three.

And lets give a little credit to Microsoft who held the shotgun on Intel
until they agreed to adopt the AMD64 instruction set. As long as AMD and
Intel are software compatible (and Intel stops playing dirty), we've got
competition, and we're all benefiting.

--
Cheers,
Jan

[Jan Klincewicz]

I am very familiar with ESXi and the difference between free and Open source.  In his post, Alan specifically referred to "open sourcing of VMware's ESX product".

That discussion had been bandied about in the press for a few years.  I was wondering if he had some new information.  Thanks.

--
Cheers,
Jan

[Rao Dronamraju]

Jim,

I have no complaints about AMD's technological prowess. I think it is an
outstanding company. I am talking more about its market (share) position.
Intel's 80% to AMD's 10.5%. It is not even a David and Goliath fight, it is
a Denise and Goliath mud wrestling!.:-)

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Jim Starkey
Sent: Thursday, November 05, 2009 3:51 PM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud
is] quite a daunting task", plus some cost data

Rao Dronamraju wrote:
>
> Possibly..but AMD is too small to make a dent on Intel at this time.
>
> AMD needs a major M&A partner..nobody is bold enough to acquire AMD
> and pissoff Intel.

> (public) cloud costs: if a unit of incremental computing capacity -
> say a single ser= ver - costs X per hour, then the equivalent capacity

[Ray DePena]

@Jan,

Intel doesn't have to be a public utility, just a monopoly.... As was pointed out, should cloud / utility take off CSPs may consider Intel a safer long term risk strategy putting more pressure on AMD and over time we may end up with all Intel clouds in a utility model....  My question is/was, in that scenario, is an Intel breakup beneficial or detrimental...

-RD

[Alan Ho]

That was my mistake. ESXi is free, not open sourced. They did open source Virtual Desktop (http://code.google.com/p/vmware-view-open-client/) which I "believe" also has binary translation (but don't quote me on it!)

What I am saying is that virtualization of a host is now pretty much for free ( e.g. ESXi ), but right now their big cash cow is vSphere - which specializes managing virtual host. As the public cloud becomes more relevant, I predict that they will move up the stack once again, and manage the virtual layer between public and private clouds. I don't think they will just "give away" their management layer within the datacenter for free, but I do predict that they will give a significant price cut to their vCloud express partners. Are there any companies on this thread that is building the layer between public & private clouds  ?

Regards,
Alan Ho

[Joseph G. Baron]

@Jan said: So does Intel have a vested interest in maintaining the status quo (server sprawl)?

 

Certainly, which is why this piqued my interest – Waxman’s views run *counter* to what you would think is Intel’s vested interest. That is why I think this is an honest assessment of private cloud from Intel-the-consumer-of-huge-amounts-of-IT-stuff, not a marketing spin from Intel-the-chipmaker. If it was a spin, you would expect Intel to say “Private clouds are great!”, because Intel would tend to benefit more from private cloud and traditional IT status quo than they would from large-scale adoption of public cloud. But Waxman says exactly the opposite, to paraphrase: “Private clouds are hard, there are no off-the-shelf [server and infrastructure] solutions”, and “Public clouds have done a very impressive job of engineering to make this all work, which would be very hard and very costly to duplicate in a private cloud”. A roundabout way of saying “Public clouds are great!”

 

BTW, if you are of the mind that cloud computing shares a lot of DNA with HPC, batch clusters, and high-throughput-computing, Intel has a great deal of experience in this area. Chip design schedules are constrained by the number of logic simulation and verification cycles you can afford to throw at the problem, and as a result, Intel has been in the high-throughput batch cluster business since, well, forever – certainly long before this kind of thing was practical on x86 architectures. So if any company on earth ought to be well-equipped to build a private cloud, it ought to be Intel. Yet Intel says building a private cloud it is a “daunting task”. Interesting…

 

Perhaps this gives some justification for the EMC / VMware / Cisco Vblock joint venture, to put the hardware and software infrastructure together in a way that is intended to make sense for cloud?

 

– Regards,

Joe Baron  +1 (919) 809-9542

[dan cox]

Intel has been a predatory business for years. AMD has been a very small
fish in this pond of about six fish: Sparc, MIPS, PowerPC, Fairchild, Intel,
and AMD. There are others but not many. The key is about $/transactions and
Intel vs. AMD is mute compared to storage $/TB, Network gear which is
overpriced for IB, 10GbE, or even Myrinet or Quadrics, needed for latency
and I/O performance.
It's like the robber barons of years ago. Intel is about to get it now after
years of monopolistic practices.

[dan cox]

In a previous email I spoke off Intel and its predatory business practices. Yes they are almost a monopoly and the Fed will break them down this time. Itanium was a three legged dog at its birth due to the shortsighted simplistic view regarding transitions from x86 instruction set or the predecessor PA/RISC either. So no apps, no storage to speak of to support a new 64 bit chip set. Wow what a shot in the foot. The arrogance and stupidity was well known by AMD hence Opteron was a success until Nehalem. 

----- Original Message -----

From: Ray DePena

To: cloud-c...@googlegroups.com

[Tarry Singh]

Intel (Capital) (even in EMEA) is suspiciously downplaying the cloud card. And I have understood from many resources that ARM is coming upstream going after the market. So Intel has a big problem coming.

Kind Regards,

Tarry Singh
CEO/Founder, Avastu
Research-Analysis-Ideation
"Start your company with your ideas. Today!"
Amsterdam, The Netherlands
Business Cell: +31630617633
Private Cell: +31629159400
LinkedIn: http://www.linkedin.com/in/tarrysingh
Blogs: http://www.ideationcloud.com
Follow me on Twitter: http://twitter.com/tarrysingh

On Wed, Nov 4, 2009 at 11:08 PM, Joseph G. Baron <jgb...@gmail.com> wrote:

Two interesting points from Intel’s Jason Waxman on private cloud and cost. He cautions that building a private cloud is “quite a daunting task”, that there is no off-the-shelf hardware and interconnect solution suitable for a private cloud. He also offers a view on conventional IT costs: servers are 50% of data center outlays, electrical power and cooling are another 23%.

 

This suggests to me a baseline comparison of traditional IT vs. (public) cloud costs: if a unit of incremental computing capacity – say a single server – costs X per hour, then the equivalent capacity in the cloud should cost at least 2X per hour. Then the other values of cloud (non-ownership, elasticity etc.) should be added *on top* of the base cloud cost.

 

The article is “Intel Details Private Cloud Hurdles”, Information Week reporting on the Cloud Computing Conference & Expo in Santa Clara.

 

(Maybe some folks on the list are there in person, but I have to be content with the trade press reports.)

[Peglar, Robert]

Intel is right - the task of building a private cloud is "daunting".
Building it optimally, that is, instead of what we mostly see in
commercial datacenters today.

The problem is not CPUs, though; the problem is storage. Most cloud
builders are locked into the mantra of rotating disk physically resident
in servers. Until they get away from that, it will remain daunting, and
act as an opposing/balancing force to economy of scale, which Rao
correctly states is a benefit of mega-centers.

Rob

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Joseph G. Baron
Sent: Thursday, November 05, 2009 9:43 PM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private
cloud is] quite a daunting task", plus some cost data

@Jan said: So does Intel have a vested interest in maintaining the
status
quo (server sprawl)?

Certainly, which is why this piqued my interest - Waxman's views run

cluster business since, well, forever - certainly long before this kind

of
thing was practical on x86 architectures. So if any company on earth
ought
to be well-equipped to build a private cloud, it ought to be Intel. Yet
Intel says building a private cloud it is a "daunting task".

Interesting.

Perhaps this gives some justification for the EMC / VMware / Cisco
Vblock
joint venture, to put the hardware and software infrastructure together
in a
way that is intended to make sense for cloud?

- Regards,

Joe Baron +1 (919) 809-9542

No virus found in this incoming message.
Checked by AVG - www.avg.com
Version: 9.0.698 / Virus Database: 270.14.52/2483 - Release Date:
11/05/09 13:52:00

[Peglar, Robert]

$/TB for storage is the wrong metric. Just like CPUs, it's about
$/transaction, and transactions/second.

Cloud builders would be well advised to use $/IOP, and (even better)
designing for optimal V, where V = CxPxR/$.
C=capacity,P=performance,R=reliability. SPC-1 gives several good clues
here.

TBs are cheap - the trick is getting to them efficiently, over a long
period of time without intervention.

Rob

[Jan Klincewicz]

@Ray:

By definition Intel is NOT a monopoly any more than Microsoft is a monopoly.  Although they may have enormous market dominance, there are other alternatives (unlike in a public utility situation where the effects are more severe.)
What do you think a breakup of Intel would accomplish ??  Sparc, Cell, etc...

I absolutely think that their actions have probably been uncompetitive and illegal (and as I mentioned they are facing stronger action in Europe than here.)  Do you think that a fragmentation of CPU instruction sets will benefit the industry.  Developers already freak about "forking" of Linux and Xen distros, even if the kernel remains unchanged.

As Dan Cox pointed out, there are certainly other platforms out there.  The market has chosen x86, which although owes nearly all of its existence to Intel, is available to others.  I am fairly certain AMD had cross-licensing agreements going way back which give them access.  As I also pointed out in another post, fragmentation of the instruction set has serious implications for software developers who in addition now to maintaining both 32 and 64-bit versions need to accommodate AMD and Intel versions of otherwise identical chips.

I don't believe Intel should be able to strong-arm and coerce (or subsidize financially) box vendors to use their products.  On the other hand, I see no legal reason whatsoever for their breakup.

<<Intel doesn't have to be a public utility, just a monopoly.... As was pointed out, should cloud / utility take off CSPs may consider Intel a safer long term risk strategy putting more pressure on AMD and over time we may end up with all Intel clouds in a utility model....  My question is/was, in that scenario, is an Intel breakup beneficial or detrimental... >>

--
Cheers,
Jan

[Ray DePena]

@Jan,

That's true, technically Intel is not a monopoly though they do own the de facto standard.  I was just reading the tea leaves a bit, if I were a public cloud provider, I'd standardize on Intel as a long term strategy mitigating the risk posed by other alternatives. 

Now if all the public cloud providers think the same way, where does that leave Intel in 5-10yrs?  I simply don't see AMD taking share from Intel as the Cloud takes off, but losing more share to Intel, and Intel is at what? 80% market share?  Does it have to be 100% before being a declarative monopoly?  At 85-95% Intel market share can AMD survive? 

[Jan Klincewicz]

@Ray,

Yes, as a matter of fact, the very definition of a monopoly is 100%.  Having SIGNIFICANT market share does not qualify.  Intel has a number of competitors in the CPU space (not just x64.)    Even there, AMD and even ARM could pose significant competitive threats if large-scale Cloud takes off because large scale Cloud providers do not care about "Intel Inside" branding.  Their CPUs are not significantly (if at all) better than those of their competitors, in fact, often they have power/heat (and pricing) issues which give them a DISadvantage in dense Data Centers.

Nothing wrong with tea leaves, but Lipton does not have a monopoly.   Anyway, what would "breaking up" Intel buy anyone ?

That's true, technically Intel is not a monopoly though they do own the de facto standard.  I was just reading the tea leaves a bit, if I were a public cloud provider, I'd standardize on Intel as a long term strategy mitigating the risk posed by other alternatives. 

Now if all the public cloud providers think the same way, where does that leave Intel in 5-10yrs?  I simply don't see AMD taking share from Intel as the Cloud takes off, but losing more share to Intel, and Intel is at what? 80% market share?  Does it have to be 100% before being a declarative monopoly?  At 85-95% Intel market share can AMD survive?

--
Cheers,
Jan

[Rao Dronamraju]

“Storage and interconnect will emerge as the critcal areas in cloud infrustructure while CPU fades in importance”

 

It is like saying hands and legs are more important than brain….did not happen so far in the human history!.

 

---

From: cloud-c...@googlegroups.com [mailto:cloud-c...@googlegroups.com] On Behalf Of Ray
Sent: Wednesday, November 04, 2009 8:49 PM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

 

Of course. Storage and interconnect will emerge as the critcal areas in cloud infrustructure while CPU fades in importance for the first phases.

 

Ray

On Nov 4, 2009, at 6:02 PM, Jan Klincewicz <jan.kli...@gmail.com> wrote:

So does Intel have a vested interest im maintaining the status quo (server sprawl) ? Would CC (public) favor less expensive AMD procs?

From: Rao Dronamraju <rao.dro...@sbcglobal.net>
Sent: Wednesday, November 04, 2009 8:10 PM
To: cloud-c...@googlegroups.com

Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

Interesting points from Jason Glassman= !. But I think he and Intel have two options.

 

1)      = Figueout a way to make the daunting tasks un-daunting in buildi= ng private clouds or

2)      = Hire a bunch of Psychologists to over come the CIO FUD! With pu= blic clouds.

 

Since Intel is almost a monopoly and h= as lot of money, it probably can do bothJ

 

From:cloud-c...@googlegroups.com [mailto:cloud-c...@googlegroups.com] = On Behalf Of Joseph G. Baron
Sent: Wednesday, November 04= , 2009 4:08 PM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data=

 

Two interesting points from Intel’s Jason Waxman on private cloud and cos= t. He cautions that building a private cloud is “quite a daunting task”, that there is no off-the-shelf hardware and interconnect solut= ion suitable for a private cloud. He also offers a view on conventional IT cost= s: servers are 50% of data center outlays, electrical power and cooling are another 23%.

 

This suggests to me a baseline comparison of traditional IT vs. (public) cloud costs: if a unit of incremental computing capacity – say a single ser= ver – costs X per hour, then the equivalent capacity in the cloud should = cost at least 2X per hour. Then the other values of cloud (non-ownership, elasti= city etc.) should be added *on top* of the base cloud cost.

 

The article is “Intel Details Private Cloud Hurdles”, Information Week reporting on the Cloud Computing Conference & Expo in Santa Clara.

[Ray DePena]

IBM doesn't have 100% (Amdahl, Fujitsu, and Hitachi) and yet....

Post Tech - DOJ Probes IBM on Mainframe Monopoly

 

Oct 8, 2009 ... The Department of Justice has begun a preliminary investigation into whether IBM has abused its monopoly in mainframe computers.

Not sure I appreciate the difference between that and Intel's position.

As to 100%.... I'm not so sure that's a requirement.

From Wikipedia:
In economics, a monopoly (from Greek monos / μονος (alone or single) + polein / πωλειν (to sell)) exists when a specific individual or an enterprise has sufficient control over a particular product or service to determine significantly the terms on which other individuals shall have access to it.^{[1][clarification needed]} Monopolies are thus characterized by a lack of economic competition for the good or service that they provide and a lack of viable substitute goods.^[2] The verb "monopolize" refers to the process by which a firm gains persistently greater market share than what is expected under perfect competition.

-RD

[Jeff Darcy]

Rao Dronamraju wrote:
>
>
> “Storage and interconnect will emerge as the critcal areas in cloud
> infrustructure while CPU fades in importance”
>
>
>
> It is like saying hands and legs are more important than brain

No, actually, it's not. Hands, legs, and brains tend to come packaged
together. Storage, interconnect, and compute come packaged separately
and can be scaled separately. The original statement might or might not
be correct*, but it can't be refuted with a bad analogy.

* Personally I think it is correct, not because differences in CPU
matter less than differences in storage or interconnects but because
more active development on the CPU side has left little room for
differentiation. There's much more room for vendors to distinguish
themselves in other areas right now, at least until the next
breakthrough in CPU speed or CPU-virtualization efficiency.

[Jan Klincewicz]

@Ray

OK .. If you want to win the argument (based on the world's foremost authority, Wikipedia), I'll gladly concede it to you, and hope that makes you happy.

<<Not sure I appreciate the difference between that and Intel's position.

As to 100%.... I'm not so sure that's a requirement.

--
Cheers,
Jan

[Rao Dronamraju]

Actually the context of the (net)conversation was nothing to do with vendors
differentiating in the clouds. The question was whether CPU will fade in the
clouds in relation to networking and storage. No it does not. Infact if you
see the developments in clouds including Google, Yahoo and Amazon, one of
the most important cloud technologies and architectures talked about is
map-reduce/Hadoop etc. Now this is clearly a CPU centric/bound technology.
With multi-cores and massive parallelism it is has been the CPU land so far.
If networking was so important in the clouds, how come we have not heard any
major technological developments in it other than regular 10Gig. Storage
probably fared better with pata byte and beyond storage scaling. In
networking we have not even heard about 100GigE and on the WAN side it seems
to be the same old T1, T2 & T3....

So lets face it, Server/CPU gets the limelight clouds or not. Because just
like brains in a body it is the most important.

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Jeff Darcy
Sent: Friday, November 06, 2009 7:29 PM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud
is] quite a daunting task", plus some cost data

[Ray DePena]

@Jan,

It's not about a debate, it's more of an assessment of Intel.  Is Intel there yet?  If not, given the cloud scenario that I posed, would it be much longer?  What would it take, 85%? 90%?  before DOJ steps in?

If not, not sure why DOJ is pursuing IBM... what's the difference there between Intel's market position in chips and IBM's in mainframes?

-RD

[Miha Ahronovitz]

Ray,

 

You are , without being aware, a Marxist.  The Monopoly is one of the few solid contributions to economics that survive time, from Karl Marx. The anti-Monopoly Law in US has its’ origin in the teachings of Karl Marx

 

“Marx's view was that capital accumulation, economies of scale, the growth of credit markets, and the dominance of the corporation in business organization would lead to the concentration and centralization of capital into fewer and fewer hands. Competition would end by destroying itself, and the large corpo­ration would assume monopoly power. With the large corporation would come a separation of ownership and control, along with a number of undesirable social consequences:

a new aristocracy of finance, a new sort of parasites in the shape of promoters, speculators, and merely nominal directors; a whole system of swindling and cheating by means of corporation juggling, stock jobbing, and stock speculation. It is private production without the control of private property.7 Possibly no other vision of the future of capitalism advanced by Marx has been more prophetic than his law of the concentration and centralization of capital”

 

http://www.economictheories.org/2008/07/karl-marx-concentration-and.html

 

Miha

From: cloud-c...@googlegroups.com [mailto:cloud-c...@googlegroups.com] On Behalf Of Ray DePena

Sent: Friday, November 06, 2009 5:08 PM
To: cloud-c...@googlegroups.com

[Ray Nugent]

Jeff, that was the point I was trying to make. Thanks for making it so much more eloquently!

Ray (the original one)

---

From: Jeff Darcy <je...@pl.atyp.us>
To: cloud-c...@googlegroups.com
Sent: Fri, November 6, 2009 5:29:07 PM

Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud is] quite a daunting task", plus some cost data

[Jeff Darcy]

Rao Dronamraju wrote:
> Actually the context of the (net)conversation was nothing to do with vendors
> differentiating in the clouds.

Please don't try to rewrite history. This thread started with Intel's
attempt at competitive positioning, and differentiation has been an
essential part of it ever since. It's not anyone else's fault if you've
failed to realize that.

> The question was whether CPU will fade in the
> clouds in relation to networking and storage.

Yes, and that can be interpreted two different ways - as a statement
about CPU's effect on system performance or as a statement about ability
to differentiate.

> No it does not. Infact if you
> see the developments in clouds including Google, Yahoo and Amazon, one of
> the most important cloud technologies and architectures talked about is
> map-reduce/Hadoop etc. Now this is clearly a CPU centric/bound technology.
> With multi-cores and massive parallelism it is has been the CPU land so far.

That's true, but has no effect on the argument at hand. Let's say, for
the sake of argument, that having CPUs that are twice as fast would
improve overall performance by 50% but having an interconnect that's
twice as fast would only improve overall performance by 25%. Thus, you
could say that CPUs are twice as important as interconnects in terms of
system performance. However, if there is no CPU twice as fast as what
your competitors but there is an interconnect three times as fast (e.g.
QDR IB vs. 10GbE) then interconnect is more important as a
differentiator. Get it?

> If networking was so important in the clouds, how come we have not heard any
> major technological developments in it other than regular 10Gig. Storage
> probably fared better with pata byte and beyond storage scaling. In
> networking we have not even heard about 100GigE and on the WAN side it seems
> to be the same old T1, T2 & T3....

Yeah, there's no such thing as IB or DWDM. Welcome to this century,
Rao. Look around a bit, there's some cool stuff.

[Jim Starkey]

Market share isn't the issue since the definition of the market can be
tailored by whim. For example, the EC has it knickers in a twist over
the "MySQL branded database market". The same technology is available,
sans brand name, from other suppliers.

If you define all microprocessors as the market, there is not shortage
of competition, except that people who buy microprocessors for server
applications generally specific AMD64 compatible processors, and for
very good reasons. Even then, also everyone knows, there is vigorous
competition from AMD. In an open market, AMD would do quite well.

And keep in mind that patents exist to grant limited time monopolies.

The issue isn't a monopoly, per se, but behaving like an monopoly.
Here, Intel almost certainly crossed the line by forcing customers away
from producing AMD based products. Anti-competitive practices aren't
defined by market share threshold, but by corporate behavior.

I applaud AMD for developing a 64 bit processor that runs 32 bit
software very well. I applaud Intel for adopting the AMD64 instruction
set to make it an industry standard. I applauded AMD for defining a
memory controller architecture consistent with high performance
contentious threads. And I applaud Intel for dumping their stupid front
side bus architecture and leaf-frogging AMD's memory architecture. In
the spirit of competition, I buy whatever is best when I need more hardware.

Isn't competition great?

Oh, IBM's problem, like Intel's, is that it is behaving like an
monopoly. The DOJ can hardly force other companies to go into business
building dinosaurs.

Now, if the EC would recognize that the "MySQL branded database" market
is a dumb concept, Oracle could get on with their rescue (and/or rape
and pillage, depending on your perspective).

Ray DePena wrote:
> @Jan,
>
> It's not about a debate, it's more of an assessment of Intel. Is
> Intel there yet? If not, given the cloud scenario that I posed, would
> it be much longer? What would it take, 85%? 90%? before DOJ steps in?
>
> If not, not sure why DOJ is pursuing IBM... what's the difference
> there between Intel's market position in chips and IBM's in mainframes?
>
> -RD
>
> On Fri, Nov 6, 2009 at 6:04 PM, Jan Klincewicz
> <jan.kli...@gmail.com <mailto:jan.kli...@gmail.com>> wrote:
>
> @Ray
>
> OK .. If you want to win the argument (based on the world's
> foremost authority, Wikipedia), I'll gladly concede it to you, and
> hope that makes you happy.
>
> <<Not sure I appreciate the difference between that and Intel's
> position.
>
> As to 100%.... I'm not so sure that's a requirement.
>
> On Fri, Nov 6, 2009 at 8:07 PM, Ray DePena <ray.d...@gmail.com
> <mailto:ray.d...@gmail.com>> wrote:
>
> IBM doesn't have 100% (Amdahl, Fujitsu, and Hitachi) and yet....>>
>
>

> Post Tech - DOJ Probes /IBM/ on /Mainframe Monopoly/
> <http://voices.washingtonpost.com/posttech/2009/10/doj_probes_ibm_on_mainframe_mo.html>
>
>
>
> Oct 8, 2009 *...* The Department of Justice has begun a
> preliminary investigation into whether /IBM/ has abused its
> /monopoly/ in /mainframe/ computers.

>
> Not sure I appreciate the difference between that and Intel's
> position.
>
> As to 100%.... I'm not so sure that's a requirement.
>

> _From Wikipedia: _
> In economics <http://en.wikipedia.org/wiki/Economics>, a
> *monopoly* (from Greek
> <http://en.wikipedia.org/wiki/Greek_language> /monos / μονος/
> (alone or single) + /polein / πωλειν/ (to sell)) exists when a

> specific individual or an enterprise has sufficient control
> over a particular product or service to determine
> significantly the terms on which other individuals shall have

> access to it.^[1]
> <http://en.wikipedia.org/wiki/Monopoly#cite_note-0>
> ^[/clarification needed
> <http://en.wikipedia.org/wiki/Wikipedia:Please_clarify>/]

> Monopolies are thus characterized by a lack of economic

> competition <http://en.wikipedia.org/wiki/Competition> for the
> good <http://en.wikipedia.org/wiki/Good_%28economics%29> or
> service <http://en.wikipedia.org/wiki/Service_%28economics%29>

> that they provide and a lack of viable substitute goods

> <http://en.wikipedia.org/wiki/Substitute_good>.^[2]
> <http://en.wikipedia.org/wiki/Monopoly#cite_note-1> The verb
> "monopolize" refers to the /process/ by which a firm gains

> persistently greater market share than what is expected under
> perfect competition

> <http://en.wikipedia.org/wiki/Perfect_competition>.

>
> -RD
>
>
> On Fri, Nov 6, 2009 at 3:25 PM, Jan Klincewicz

> <jan.kli...@gmail.com <mailto:jan.kli...@gmail.com>>

> <mailto:ray.d...@gmail.com>> wrote:
>
> @Jan,
>
> Intel doesn't have to be a public utility,
> just a monopoly.... As was pointed out, should
> cloud / utility take off CSPs may consider
> Intel a safer long term risk strategy putting
> more pressure on AMD and over time we may end
> up with all Intel clouds in a utility
> model.... My question is/was, in that
> scenario, is an Intel breakup beneficial or
> detrimental...
>
> -RD
>
> On Thu, Nov 5, 2009 at 3:55 PM, Jan Klincewicz
> <jan.kli...@gmail.com

> [mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>]
> *On Behalf Of *Jan Klincewicz
> > *Sent:* Wednesday, November 04, 2009
> 8:03 PM
> > *To:*
> cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>

> > *Subject:* [ Cloud Computing ] Re:
> Intel's Waxman: "[building a
> > private cloud is] quite a daunting
> task", plus some cost data
> >
> > So does Intel have a vested interest
> im maintaining the status quo
> > (server sprawl) ? Would CC (public)
> favor less expensive AMD procs?
> >
> >
> ------------------------------------------------------------------------
> >
> > *From: *Rao Dronamraju
> <rao.dro...@sbcglobal.net

> <mailto:rao.dro...@sbcglobal.net>>

> > *Sent: *Wednesday, November 04, 2009
> 8:10 PM
> > *To:
> *cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>

> > *Subject: *[ Cloud Computing ] Re:
> Intel's Waxman: "[building a
> > private cloud is] quite a daunting
> task", plus some cost data
> >

> > Interesting points from Jason
> Glassman= !. But I think he and Intel
> > have two options.
> >
> > 1) = Figueout a way to make the
> daunting tasks un-daunting in buildi=
> > ng private clouds or
> >
> > 2) = Hire a bunch of Psychologists
> to over come the CIO FUD! With pu=
> > blic clouds.
> >
> > Since Intel is almost a monopoly and
> h= as lot of money, it probably
> > can do bothJ
> >
> >
> ------------------------------------------------------------------------
> >
> >
> *From:*cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >

> [mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>]
> = *On Behalf Of *Joseph G. Baron
> > *Sent:* Wednesday, November 04= ,
> 2009 4:08 PM
> > *To:*
> cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>

> <mailto:jgb...@gmail.com>

[Ray DePena]

Miha,

I'm not aware of all of Marx's views, but you can find any person's views and I'm sure there would be areas where we agree and areas where we disagree.

In this particular area, it doesn't require a leap of imagination to see that a monopoly is counter to the consumer benefits derived from capitalistic competition.

You may find it interesting that Adam Smith, the father of capitalism, did not trust businessmen.  I agree with much of what Smith has written, as opposed to much of the hyperbole attributed to Smith by today's political demagogues.

It doesn't make me a "Smither" anymore than a "Marxist". 

All that said, your observation is an interesting one, perhaps one day I'll get around to reading about Marx.  At the moment, I'm tied up with Kant, Wittengstein, Russell, and Moore. :-)

-RD

[Rao Dronamraju]

"That's true, but has no effect on the argument at hand. Let's say, for the
sake of argument, that having CPUs that are twice as fast would improve
overall performance by 50% but having an interconnect that's twice as fast
would only improve overall performance by 25%. Thus, you could say that
CPUs are twice as important as interconnects in terms of system performance.
However, if there is no CPU twice as fast as what your competitors but there
is an interconnect three times as fast (e.g.
QDR IB vs. 10GbE) then interconnect is more important as a differentiator.
Get it?"

This is like saying, you are going to a singing contest and you are
differentiating yourself by dressing differntly. Nobody cares how you are
dressed (LAN bandwidth), they care your singing ability (CPU/Processing
power) even if the singing abilities are pretty close.

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Jeff Darcy
Sent: Saturday, November 07, 2009 6:14 AM
To: cloud-c...@googlegroups.com
Subject: [ Cloud Computing ] Re: Intel's Waxman: "[building a private cloud
is] quite a daunting task", plus some cost data

[Jan Klincewicz]

@Ray:

Initially, my argument was about defining Intel as a MONOPOLY, which, again, by definition, it is not.  I hate to dishonour the Latins after they worked so hard for millennia to leave future generations and cultures  roots which could be adapted to their disparate languages in order to add precise meaning to words.  MONO, meaning  ONE means .... 1.  In the case of the word Monopoly there is ONE seller and MANY buyers.  This is not subject to debate or interpretation.  I cannot (in all honesty) tell my wife I am monogamous if I am nailing other women 20% of the time. 

There are sufficient words left to describe Intel, but even DUOpoly does not reflect their position as accurately as OLIGopoly as there are more than one entrant in each of their markets (regardless of relative position.)

I really hate to split linguistic hairs in a discussion, but unless we can find agreement on definitions of common words, arguing the implications of those words becomes pointless.

The lede you quote below "The Department of Justice has begun a preliminary investigation into whether IBM has abused its monopoly in mainframe computers." is inaccurate (wrong).   The DOJ is looking into anti-competitive behavior by IBM under the U.S. Anti-Trust statutes to help ensure that IBM does not BECOME a monopoly.  Neither Wikipedia not journalists are gospel. Without an EXCLUSIVE position, Intel does not have a monopoly. So to answer your question, there is essentially no difference between Intel's position in chips, and IBM's in mainframes.  Neither has a monopoly, and both are under Anti-trust investigations by the DOJ (as well as other bodies internationally.)

That being said, I think it is clear that Intel has frequently been at least accused of anti-competitive behavior.  Should they be punished for it ?  I think so.  Should / Can the DOJ "break them up ?"  The question is, into what ? Below are their product divisions.  Beyond that they do pure research (which helped develop standards like USB, PXE, Wake on Lan, Serial ATA etc.)  So how would you suggest breaking them and, and to what end and whose benefit ?

    * Processors
    * Motherboards
    * Chipsets
    * Desktop
    * Notebook
    * Business PCs
    * Server
    * Workstation
    * Intel® Graphics Technology
    * Embedded & Communications
    * Solid-State Drives and Caching
    * Storage
    * Consumer electronics
    * Software
    * Intel® Health
    * Ethernet

--
Cheers,
Jan

[Jan Klincewicz]

BTW, I do have an error to correct.  It is the FTC who began investigations into Intel's behavior and not the DOJ.    According to an article in Network World

"The US DOJ did not bring a single antitrust case against a dominant company during the 8 years of George W. Bush's administration. Instead, the DOJ tried to build a "safe harbor of conduct" for those dominant firms to protect them from antitrust consideration." "Intel was already fined $1.44 billion earlier this year by the EU for anti-competitive behavior against AMD, and $25 million in South Korea last year."

--
Cheers,
Jan

[Jan Klincewicz]

@ Jim

Interesting points, but still ..

<Oh, IBM's problem, like Intel's, is that it is behaving like an monopoly.  The DOJ can hardly force other companies to go into business building dinosaurs>>

There is no law again BEHAVING like a monopoly (but when a company does so, it is only fooling itself.)  The Anti-trust laws were designed to prevent companies from actually BECOMING monopolies.
Anti-competitive behavior (such as bullying suppliers or customers so they will not deal with your competitors) is such a behavior.  Also, using a superior financial position to sell parts below cost for a
period sufficient to bankrupt a competitor is illegal.

Acting in an arrogant manner to partners and customers, failing to improve products or innovate, charging exorbitant prices, etc. are all monopolistic behavior, and are fully legal (if ill-advised.)

--
Cheers,
Jan

[Jim Starkey]

Jan Klincewicz wrote:
> @ Jim
>
> Interesting points, but still ..
>
> <Oh, IBM's problem, like Intel's, is that it is behaving like an
> monopoly. The DOJ can hardly force other companies to go into
> business building dinosaurs>>
>
> There is no law again BEHAVING like a monopoly (but when a company
> does so, it is only fooling itself.) The Anti-trust laws were
> designed to prevent companies from actually BECOMING monopolies.

"Behaving like a monopoly" is not a well defined term, so let's not
waste time trying nail it down.

To quote from the ubiquitous Wikipedia:

In many jurisdictions, competition laws </wiki/Competition_law>
place specific restrictions on monopolies. Holding a dominant
position or a monopoly in the market is not illegal in itself,
however certain categories of behaviour can, when a business is
dominant, be considered abusive and therefore be met with legal
sanctions. A government-granted monopoly
</wiki/Government-granted_monopoly> or /legal monopoly/, by
contrast, is sanctioned by the state, often to provide an incentive
to invest in a risky venture or enrich a domestic constituency
</wiki/Interest_group>. The government may also reserve the venture
for itself, thus forming a government monopoly
</wiki/Government_monopoly>.

There is no way to prevent monopolies, not should there be. If I invent
and patent a better mousetrap, I have, by act of the USPTO, itself
performing a Constitutional duty, a monopoly on that mousetrap. Nothing
wrong there. And I can charge as much as I'd like. If I tell my
customers that if they want to buy my mousetrap, they have to sever
business relationships with my competitors, then I'm abusing my position
and can expect to suffer.

Intel does not have a monopoly on the x86_64 market, let alone the
microprocessor market. The lack of a monopoly, however, doesn't grant
them the right to act in an anti-competitive manner.

How did we ever get here? Does this have anything even remotely to do
with cloud computing?

Jan, have your last say, then let's let this drop.

> <mailto:jan.kli...@gmail.com

> <mailto:jan.kli...@gmail.com>>> wrote:
> >
> > @Ray
> >
> > OK .. If you want to win the argument (based on the world's
> > foremost authority, Wikipedia), I'll gladly concede it to
> you, and
> > hope that makes you happy.
> >
> > <<Not sure I appreciate the difference between that and Intel's
> > position.
> >
> > As to 100%.... I'm not so sure that's a requirement.
> >
> > On Fri, Nov 6, 2009 at 8:07 PM, Ray DePena
> <ray.d...@gmail.com <mailto:ray.d...@gmail.com>

> > <mailto:ray.d...@gmail.com <mailto:ray.d...@gmail.com>>>

> <mailto:jan.kli...@gmail.com> <mailto:jan.kli...@gmail.com

> <mailto:ray.d...@gmail.com <mailto:ray.d...@gmail.com>>> wrote:
> >
> > @Jan,
> >
> > That's true, technically Intel is not a monopoly
> > though they do own the de facto standard. I was
> just
> > reading the tea leaves a bit, if I were a public
> cloud
> > provider, I'd standardize on Intel as a long term
> > strategy mitigating the risk posed by other
> > alternatives.
> >
> > Now if all the public cloud providers think the same
> > way, where does that leave Intel in 5-10yrs? I
> simply
> > don't see AMD taking share from Intel as the Cloud
> > takes off, but losing more share to Intel, and Intel
> > is at what? 80% market share? Does it have to
> be 100%
> > before being a declarative monopoly? At 85-95%
> Intel
> > market share can AMD survive?
> >
> > On Fri, Nov 6, 2009 at 4:49 AM, Jan Klincewicz
> > <jan.kli...@gmail.com
> <mailto:jan.kli...@gmail.com>

> > <mailto:jan.kli...@gmail.com

> > <mailto:ray.d...@gmail.com

> <mailto:ray.d...@gmail.com>>> wrote:
> >
> > @Jan,
> >
> > Intel doesn't have to be a public utility,
> > just a monopoly.... As was pointed out,
> should
> > cloud / utility take off CSPs may consider
> > Intel a safer long term risk strategy
> putting
> > more pressure on AMD and over time we
> may end
> > up with all Intel clouds in a utility
> > model.... My question is/was, in that
> > scenario, is an Intel breakup beneficial or
> > detrimental...
> >
> > -RD
> >
> > On Thu, Nov 5, 2009 at 3:55 PM, Jan
> Klincewicz
> > <jan.kli...@gmail.com
> <mailto:jan.kli...@gmail.com>

> > <mailto:jan.kli...@gmail.com

> > <mailto:jsta...@nimbusdb.com

> [mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>
> >
> <mailto:cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>>]
> > *On Behalf Of *Jan Klincewicz
> > > *Sent:* Wednesday, November
> 04, 2009
> > 8:03 PM
> > > *To:*
> > cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >

> <mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>>
> > > *Subject:* [ Cloud Computing ] Re:
> > Intel's Waxman: "[building a
> > > private cloud is] quite a daunting
> > task", plus some cost data
> > >
> > > So does Intel have a vested
> interest
> > im maintaining the status quo
> > > (server sprawl) ? Would CC
> (public)
> > favor less expensive AMD procs?
> > >
> > >
> >
> ------------------------------------------------------------------------
> > >
> > > *From: *Rao Dronamraju
> > <rao.dro...@sbcglobal.net
> <mailto:rao.dro...@sbcglobal.net>
> >

> <mailto:rao.dro...@sbcglobal.net

> <mailto:rao.dro...@sbcglobal.net>>>
> > > *Sent: *Wednesday, November
> 04, 2009
> > 8:10 PM
> > > *To:
> >
> *cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >

> <mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>>
> > > *Subject: *[ Cloud Computing ] Re:
> > Intel's Waxman: "[building a
> > > private cloud is] quite a daunting
> > task", plus some cost data
> > >

> > > Interesting points from Jason
> > Glassman= !. But I think he and
> Intel
> > > have two options.
> > >
> > > 1) = Figueout a way to make the
> > daunting tasks un-daunting in
> buildi=
> > > ng private clouds or
> > >
> > > 2) = Hire a bunch of Psychologists
> > to over come the CIO FUD! With pu=
> > > blic clouds.
> > >
> > > Since Intel is almost a
> monopoly and
> > h= as lot of money, it probably
> > > can do bothJ
> > >
> > >
> >
> ------------------------------------------------------------------------
> > >
> > >
> >
> *From:*cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >
> <mailto:cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>>
> > >
> >

> [mailto:cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >
> <mailto:cloud-c...@googlegroups.com

> <mailto:cloud-c...@googlegroups.com>>]
> > = *On Behalf Of *Joseph G. Baron
> > > *Sent:* Wednesday, November 04= ,
> > 2009 4:08 PM
> > > *To:*
> > cloud-c...@googlegroups.com
> <mailto:cloud-c...@googlegroups.com>
> >

> <mailto:cloud-c...@googlegroups.com

> > <mailto:jgb...@gmail.com

[Ray DePena]

@Jim, re:

"How did we ever get here?  Does this have anything even remotely to do
with cloud computing?"

The relationship to CC is a school of thought / view that large CSPs may standardize on Intel increasing their market share beyond the current 80%, so I posed the question as to whether AMD and others could survive in a market (where FAB investments are significant) with 5% or less share of the market, and if not, would that leave Intel as a monopoly.  If so, whether they would suffer the same fate as AT&T's breakup.  Pros and Cons of such, etc.

-RD

[Jeff Darcy]

Rao Dronamraju wrote:
> This is like saying, you are going to a singing contest and you are
> differentiating yourself by dressing differntly. Nobody cares how you are
> dressed (LAN bandwidth), they care your singing ability (CPU/Processing
> power) even if the singing abilities are pretty close.

Enough with the broken analogies, already. I've personally worked on
dozens of systems and applications where network bandwidth - and even
more often network latency - affected overall performance as much or
more than CPU power did. It's part of the competition; "only CPU
matters" is a hopelessly narrow-minded view that will not serve anyone
well in the real world, and I have nothing else to say to anyone so
misguided.