Perspectives - Private Clouds Are Not The Future

Sunday, January 17, 2010

Cloud computing is an opportunity to substantially improve the economics of enterprise IT. We really can do more with less.

I firmly believe that enterprise IT is a competitive weapon and, in all industries, the leaders are going to be those that invest deeply in information processing. The best companies in each market segment are going to be information processing experts and because of this investment, are going to know their customer better, will chose their suppliers better, will have deep knowledge and control of their supply chains, and will have an incredibly efficient distribution system. They will do everything better and more efficiently because of their information processing investment. This is the future reality for retail companies, for financial companies, for petroleum exploration, for pharmaceutical, for sports teams, and for logistics companies. No market segment will be spared and, for many, it’s their reality today. Investment in IT is the only way to serve customers and shareholders better than competitors.

It’s clear to me that investing in information technology is the future of all successful companies and it’s the present for most. The good news is that it really can be done more cost effectively, more efficiently, and with less environmental impact using cloud computing. We really can do more with less.

The argument for cloud computing is gaining acceptance industry-wide. But, private clouds are being embraced by some enterprises and analysts as the solution and the right way to improve the economics of enterprise IT infrastructure. Private clouds may feel like a step in the right direction but scale-economics make private clouds far less efficient than real cloud computing. What’s the difference? At scale, in a shared resource fabric, better services can be offered at lower cost with much higher resource utilization. We’ll look at both the cost and resource utilization advantages in more detail below.

At very high-scale it’s both affordable and efficient to have teams of experts in power distribution and mechanical systems on staff. The major cloud computing providers have these teams and are inventing new techniques to lower costs, improve efficiency, and provide more environmentally sound solutions. This is very hard to do cost effectively at scale of less than 10s of megawatts. Continuing that same argument to other domains, cloud computing providers have teams specialized in server and storage design. And they are deeply invested in networking gear hardware and software. All of this is hard to justify at private cloud scales.

Cloud computing providers have 24x7 staff to monitor the services and to respond to customer issues. Doing service monitoring right is incredibly difficult and I’ve never seen it done well at anything less than multi-megawatt scales.

Cloud computing providers have some of the best distributed systems specialists in the world. They also have open source experts and depend deeply upon both open source and internally produced software. They do this for two reasons: 1) at high-scale, things fail in new and interesting ways – operational excellence only comes from intimate knowledge of the entire hardware and software stack, and 2) when running at the high scale needed for efficiency, software licensing costs give up much of the excellent economics of a cloud service.

Resource utilization is even a stronger argument to move to a high-scale, shared infrastructure cloud. At scale, with high customer diversity, a wonderful property emerges: non-correlated peaks. Whereas each company has to provision to support their peak workload, when running in a shared cloud the peaks and valleys smooth. The retail market peaks in November, taxation in April, some financial business peak on quarter ends and many of these workloads have many cycles overlaid some daily, some weekly, some yearly and some event specific. For example, the death of Michael Jackson drove heavy workloads in some domains but had zero impact in others. A huge eastern seaboard storm drives massive peaks in a few businesses but has no impact on most. Large numbers of diverse workloads tend to average out and yield much higher utilization levels than are possible at low scale. Private clouds can never achieve the utilization levels of shared clouds.

Last week Alistair Croll wrote an excellent InformationWeek article arguing that “the true cloud operators will have an unavoidable cost advantage because it's all they worry about. They'll also be closer to consumers (because they have POPs everywhere and partnerships with content delivery systems), and connecting with consumers and partners will become an increasingly essential part of any enterprise IT strategy.” Have a look at Private Clouds are a Fix, Not the Future.

Private clouds are better than nothing but an investment in a private cloud is an investment in a temporary fix that will only slow the path to the final destination: shared clouds. A decision to go with a private cloud is a decision to run lower utilization levels, consume more power, be less efficient environmentally, and to run higher costs.

James Hamilton

e: jrh@mvdirona.com

w: http://www.mvdirona.com

b: http://blog.mvdirona.com / http://perspectives.mvdirona.com

Sunday, January 17, 2010 9:20:10 AM (Pacific Standard Time, UTC-08:00)

Comments [24] - Trackback
Services

Sunday, January 17, 2010 11:56:23 AM (Pacific Standard Time, UTC-08:00)

While I certainly agree that private clouds are not a tenable long-term solution and that the large-scale providers have a definite edge in cost, I think that data lock-in is still a fundamental concern to many IT departments. If a business's value is truly its data and that data is very large, as it certainly will be, what will be the driver to push the movement of that data onto infrastructure controlled by others where moving between providers is non-trivial?

One possibility is that as network bandwidth increases movement of multi-TB or PB datasets becomes somewhat more feasible, but I seriously doubt that network bandwidth will keep up with the explosion in data. From the provider's perspective it would seem that once you can get an enterprise to put its core data in your cloud you have them long term since we know that the location of data usually drives the location of computing. That could be a very uncomfortable position for an enterprise--lack of options long-term.

What are your thoughts on this? Is it just a matter of IT cost-competitiveness such that companies will have no choice in order to stay competitive? Is the data lock-in concern merely a knee-jerk reaction to the unfamiliar?

Zach Hill

Sunday, January 17, 2010 12:01:41 PM (Pacific Standard Time, UTC-08:00)

I am in absolute agreement James. The vision expressed here, in books like The Fourth Paradigm and in other blogs like IBM's A Smarter Planet show the impact that huge, interconnected computing resources will have on the world. Harnessing, correlating and aggregating information from the cloud will morph from advantage to necessity as the impact of these technologies spreads further into the physical world. Growing constraints in resource availability per capita will effectively force the world to embrace efficiency in order to sustain and grow quality of life, etc. We are truly just at the dawn of the Information Age.

What an exciting time to be involved in this industry.

-Tom

Tom

Sunday, January 17, 2010 7:25:44 PM (Pacific Standard Time, UTC-08:00)

I'd like to disagree. One of my acquaintances works for a department of defense ( more than one department by that name in the world.)
They are keen to use clouds internally for their own systems (where it has huge potential), but would never ever use an external cloud.

The world is not entirely composed of e-commerce companies. Some entities ( especially in government-defense) can't share network and resources with others for reasons that no amount of job and communications encryption can ever cover.

Tim Williscroft

Sunday, January 17, 2010 10:18:12 PM (Pacific Standard Time, UTC-08:00)

Tim has a good point in the comments above that some companies many need private clouds for privacy and security. The premium private clouds require may be justified by that benefit.

There is another point that you don't quite get to, James, that, since scale provides lower costs and higher utilization, only the biggest cloud computing providers will survive. And, for those that do survive, margins are likely to be very tight.

Greg Linden

Monday, January 18, 2010 5:35:54 AM (Pacific Standard Time, UTC-08:00)

Greg,I think you are right that the cloud market is going to be based upon thin margins at very high volume. Any cloud provider expecting to find software-like margins is on the wrong path. I love business models that are good for customers and still support the margins needed to re-invest in the offering.

Tim you may be right that there are exceptions like defense with infinite capital that will chose to pay more. Many of the same companies that often get pointed to as unable to use services already do. Most financial services companies use the ADP payroll service. Most use separate security services even though physical security is the bedrock upon which all other security is built. Without physical security there is nothing. Companies are using services, can use services, and this use of services will expand. I haven't a clue what the department of defense will do but let's not use them as a predictor of which direction the rest of the computing world is heading. They are a wonderful customer but atypical and not particularly predictive of market trends.

James Hamilton
jrh@mvdirona.com

James Hamilton

Monday, January 18, 2010 5:55:24 AM (Pacific Standard Time, UTC-08:00)

Good hearing from you Zach. I think you are right that the potential of data lock-in needs to be addressed in any thorough discussion of cloud computing. I 100% agree. If customers aren't convinced they can move their data efficiently, its a big issue especially given that data sizes are growing much faster than WAN bandwidth costs are falling.

The economic solution to this one is the modern version of Tanenbaum's "station wagon full of tapes". Cheap, mult-terabyte, USB-connected disk drives sent via courier. The AWS Inport/Export service (http://aws.amazon.com/importexport/) is one example of such a service.

Allowing efficient data migration is super important. If the workload is whatever you can run in generic Linux (or Windows) instances and the data can move, there is no data lock-in. This is one of the reasons that Greg is right above where he predicts thin margins in cloud offerings. There won't be the lock-in to support unreasonable pricing.

Thanks for the comment Zach.

James Hamilton
jrh@mvdirona.com

James Hamilton

Monday, January 18, 2010 5:57:49 AM (Pacific Standard Time, UTC-08:00)

Tom, I agree with your point that its an exciting time to be in our industry. Things are changing quickly and the changes are big. 10 years ago I thought things were moving quickly but the amplitude of the change was far less. Its an exciting time.

James Hamilton
jrh@mvdirona.com

James Hamilton

Monday, January 18, 2010 11:29:07 AM (Pacific Standard Time, UTC-08:00)

I like the general idea that "private clouds" are largely a fool's errand for many organizations, but one thing that makes me nervous is the notion of a couple entities having unilateral control (cost, API, functionality, etc.) over what is quickly becoming the platform for most new developments in large-scale computing. We've talked before about how difficult it is to move from one cloud provider to another (moving data, changing APIs, etc.) - these are not insignificant barriers, so cloud providers must carry the torch on behalf of their customers' best interest.

While "public" cloud computing is obviously cost- and resource-efficient, it also introduces the potential for anti-competitive behavior. I'm not suggesting that this is happening or would even be intended by the major players, but certainly others will (and probably already have begun to) make that allegation at some point.

While many things in the cloud are different than the OS, browser and framework battles of yesteryear, there are also a number of similarities. I'm interested to see how things turn out, and how competition and innovation will evolve on this playing field.

Andrew Cencini

Monday, January 18, 2010 11:39:37 AM (Pacific Standard Time, UTC-08:00)

This also feels to me like something that would make a great (and timely) debate for CACM via the most recent editor's letter, "More Debate Please". (http://cacm.acm.org/magazines/2010/1/55739-more-debate-please/fulltext).

Andrew Cencini

Wednesday, January 20, 2010 2:42:25 PM (Pacific Standard Time, UTC-08:00)

Andrew, I understand your concern on the possibility of a small number of competitors supporting higher prices but I think its highly probable that this market segment won't settle down to 1, 2, or 3 players. Most likely the winners will be fairly large but I'm confident there will be many winners. Expect 10s. Probably not 100s but definitely not 2 or 3.

This market is going to stay hungry and aggressive and keep doing good things for customers.

James Hamilton
jrh@mvdirona.com

James Hamilton

Wednesday, January 20, 2010 4:17:47 PM (Pacific Standard Time, UTC-08:00)

I expect that many companies today will not be able to make the leap from their current architectures and embrace a public cloud. This creates a market inefficiency that the next wave of startups may be able to exploit to great pricing advantages against other market participants.

I think one thing that makes the lockin problem different this time around is pricing transparency.

Steve

Steve Severance

Thursday, January 21, 2010 7:05:19 AM (Pacific Standard Time, UTC-08:00)

Great observation Steve. Super interesting.

James Hamilton
jrh@mvdirona.com

James Hamilton

Thursday, January 21, 2010 12:38:56 PM (Pacific Standard Time, UTC-08:00)

Love your blog.

Private clouds soothe the anxiety enterprises may have over loss of control in a public cloud?

I see an analogy to the tendency of hoarding cash in mattresses until banks proved themselves. Private clouds will possibly ease the eventual acceptance and the journey to public clouds.

Do you think that large enterprises would want to jump onto the service provider bandwagon to reap scale benefits internally for their private clouds?

Regards
Sameer

Sameer Deokule

Thursday, January 21, 2010 6:29:28 PM (Pacific Standard Time, UTC-08:00)

Sameer asked: "Do you think that large enterprises would want to jump onto the service provider bandwagon to reap scale benefits internally for their private clouds?"

I suppose its possible but one would think that as soon as a company figures out that private is more expensive, then starts selling externally to get the scale benefits, the most likely place for them to end up is as a real (rather than private) cloud provider.

I agree with your analogy. Most companies use services heavily: payroll, security, management consulting, HR consulting, etc.

James Hamilton
jrh@mvdirona.com

James Hamilton

Saturday, January 23, 2010 5:43:25 PM (Pacific Standard Time, UTC-08:00)

I'm interested to see how the Cloud evolves in the next decade. Will we see a utility model, just like electricity, where you simply upload an application and let it "do its thing?" The concept of the individual computer would completely abstracted away, and replaced with "processing seconds" or some other measure.

It's partially what our startup is banking on, but a good mental exercise nonetheless.

Bradford

Sunday, January 24, 2010 1:49:50 AM (Pacific Standard Time, UTC-08:00)

Hello Mr. Hamilton,

GREAT article and so right. However, I see one topic not addressed thoroughly in the article nor the comments: security. It's true public clouds have the advantage of scale and can therefor do their different services way much better than smaller scale infrastructures, being clouds or not. The same premise holds for security.

True that most security threats at smaller scale come from inside the enterprise confines. However, one can not deny that a hugh public cloud infrastructure is a vastly more attractive target for external threats than those small-scale scattered interconnected computersystems spread all over the world. What if compromising one virtual node in a huge cloud compromises the whole ... Physical node? Rack? Cluster? Datacenter? Cloud?

As I read most objections, it is the impact this insecurity, rather uncertainty, has on the operationality of cloud-hosted infrastructure: Is my data really safe? And, do I have good availability? Actually, the thing is that in cloud environment, all things have to be thought off and done differently. On both domains. What are the blueprints here?

I think that is where the public cloud providers will have to convince the naysayers.

Wim Van Leuven

Sunday, January 24, 2010 2:03:15 AM (Pacific Standard Time, UTC-08:00)

I'm not on the same line as other commenters who are saying enterprises fear to loose control. Most enterprise these days are accustomed to the idea of outsourcing, I would thinking, and to managing these partnerships to a level that fits their bill.

Heck, most of them already outsource their infrastructure to classical datacenters. Those datacenters are mostly threatened by those few large cloud players that are picking on their market share. Combined with those cloud providers and other large players offering standardized applicative services like hosted mail, groupware, etc, they are anxiously looking for new market opportunities, preferably in the market of cloud services.

I might even suppose that A-type server hardware vendors are anxiously looking at those large datacenters, because they might be forcing them into a disastrous price competition as a few singular customers arise on the playing field.

All these different players are of course touting the private cloud idea at these keeps their customers closer to their side than driving them to the small pack of wolves that seem to be doomed to bind them all. The one ring?

Wim Van Leuven

Sunday, January 24, 2010 2:53:04 AM (Pacific Standard Time, UTC-08:00)

Hybrib clouds, that is private clouds that use public clouds when required are probably going to fill the gap for quite sometime.

Data that cannot legally be kept in unknown locations, security, and lock-in are still likely to stop full adoption to public clouds for sometime. But no one is going to increase local capacity just to host a company blog.

Eventually computing will be a pure utility, like water or power, but that is still a decade away.

Sunday, January 24, 2010 6:34:32 AM (Pacific Standard Time, UTC-08:00)

Thanks for the comments Wim, Bradford, and DE. I enjoyed reading through them this morning. Wim, you were arguing essentially that "perception is reality" and enterprise customers needs to be convinced that the systems are secure and will be highly available. There is no question that is true. DE made a similar argument saying the move to public clouds will happen but it'll take a decade. Past major architecturally transitions re-enforce that position. It really does take time for the industry to step through a generational change. IBM mainframes (Z-Series) continue to run some workloads although it is a rapidly shrinking fraction of the worlds computing base. Nonetheless they are still selling and, last I checked, sales were increasing even though clearly overall market segment share was decreasing.

Big changes take time. Given the saving to be had and the reduced environmental impact to be achieved through more efficient operation and higher utilization, we need to find a way to make this transition more quickly. Ideas?

James Hamilton
jrh@mvdirona.com

James Hamilton

Monday, January 25, 2010 12:14:41 PM (Pacific Standard Time, UTC-08:00)

Hello James,

"perception is reality" ... I like that. However, my point is also that public clouds have history, habits, establishment and suppliers/competitors making the status quo. So it is indeed a hard fight.

As a matter of fact, I'm not sure how to tackle it properly. Blue prints? Thorough support and screening of early adopter setups by some architectural review board? Also make use of and support 'early consultant' to evangelize?

Wim Van Leuven

Monday, February 01, 2010 12:27:22 AM (Pacific Standard Time, UTC-08:00)

I had a knee-jerk reaction to your post and wanted to post why I disagreed about it. It took me an hour to figure out that I was debating the wrong reasons. Then yesterday I found the real reasons in a slide deck from Craig Balding (http://www.slideshare.net/craigbalding/the-belgian-beer-lovers-guide-to-cloud-security) which has more realistic down to earth reasons why larger organizations are slightly scared of it today. They boil down to a few points

* Loosing control of infrastructure, security, decision making power
* Less transparency on how infrastructure is managed/controled
* Migrating existing applications might be expensive
* Platform lock-in is scary (more for PAAS based clouds)
* Legal/Compliance/Search-seizure rules
* Billing-fraud/CPU-theft-detection
* Better way of allowing developers (internal users) to control cloud instances. AD integration of some kind would be nice.

In general I still agree with your conclusion and I think its a matter of time.... in some way this is no different than organizations switching to internet from the pre-internet periods to do their business.

rkt

Royans

Monday, February 01, 2010 5:35:46 AM (Pacific Standard Time, UTC-08:00)

In the comment above, rkt argues that there are many reasons "why larger organizations are slightly scared of shared clouds] today" and then concludes "n general I still agree with your conclusion and I think its a matter of time."

It sounds like rkt and are in agreement on the overall direction but may disagree on timing. With technology shifts, timing is usually where the best debates can be had. In favor of a quick shift, we have a cost advantage of several magnitudes. Small percentage difference technology shifts typically take a long time but the really high impact cost shifts tend to happen more rapidly. Looking at the other side, arguing for a slow shift, we have the complexity of large enterprise IT infrastructure and the difficulty of moving it quickly regardless of the gains. Both good arguments in my view.

There are a couple of factors we should keep in mind as we think through the likely timing of this shift that most of us agree is inequitable. The first is that when the cost shift is huge, the move can be quick. 20 years ago most enterprises wrote there own finance, HR, and customer relationship software. The cost difference of buying from SAP, Baan, Peoplesoft, Seibel was so huge when compared to internally developed systems, this transition happened incredibly quickly. What this proves is that the enterprise can move very quickly when the cost difference is huge. Quick shifts can happen although we know they mostly don't.

Generally rkt and I agree on the outcome and we're both thinking through the timing of the move. I'm arguing that the cost differences are enough to provoke a faster than typical transition but I also argue that moving an enterprises complete IT infrastructure in a single move would be foolish and near impossible. The move will start with new applications and applications that either bring low risk or are easier to move. As companies begin to get experience from this small transitions, they will continue to move more but it will take time and won't happen over night.

I would argue against a plan to move everything in a single operation but I would also strongly argue that any company that doesn't have at least a pilot public cloud project underway is making a big mistake. And, if the goal is to get to a public cloud, I question the value of going through all the cost, risk, and complexity of moving to a cloud to get to a private cloud expecting to later move again. Its just too hard and too expensive to plan to move the infrastructure twice.

James Hamilton
jrh@mvdirona.com

James Hamilton

Thursday, February 04, 2010 7:11:58 PM (Pacific Standard Time, UTC-08:00)

Great blog and discussion. Another factor: To make financial sense, private clouds need not cost less to build and run than public clouds - only less than public clouds PLUS MARKUP.

I haven't analyzed provider pricing, though I've heard of decreases and increases, such as EC2's plan to end free inbound traffic. My wholly subjective guesstimate (please correct me) is that the unit price of robust computing in the public cloud has been dropping no faster than maybe 5 or 8% per year. Unlike Windows laptops and flatscreen TVs.

So two things might be happening. With architects like you, public providers may be riding a black diamond cost slope, maybe 10 or 25% per year, and widening their margins. Or, costs are falling slowly, and providers are maintaining moderate margins to grow the market. More likely, in between.

In the first case, it won't take much leakage of secret sauces and top talent from proprietary "public" clouds for private clouds to undersell proprietary clouds plus markups. Hardware makers and independent consultants have incentives to accelerate the leakage.

In the second case, where the cost curve at even the best public provider slopes gently downward, the private cloud manager can relax, knowing that one or two years' technology advantage on the public side will be erased by the inevitable markup.

Yes, the cross-customer averaging of load peaks is a strong pro-public argument. But don't some websites find it cheaper to run their own servers for baseline loads, and pay the public cloud only for surges? Lower cost + higher efficiency = happiness!

Finally, your recent response questioning the value of moving to a private cloud only to move again seems slightly two-edged: if clouds don't lock you in, why should the second move be hard?

Paul N

Tuesday, February 09, 2010 5:45:12 AM (Pacific Standard Time, UTC-08:00)

Paul, you raise three important points/questions: 1) is public+profit < private, 2) if private is 2 years behind won't that advantage get swallowed by the "+ profit" component of a public cloud, and 3)if the first move is possible then why isn't the second move cheap -- essentially what's the disadvantage of making a private cloud an intermediate resting point?

On the first point, is public+profit<private, I'm arguing two points: 1) the advantage of a public cloud are large and 2) the expected outcome is a very large number of customers with a very small profit margins. Essentially, I'm arguing the cloud computing business looks a lot more like retail than the crazy 80% margins we see in software (and advertising) today. Actually this isn't a very prescient prediction since its largely true today. Margins will be small and volumes will be prodigious making it a perfectly good business both good for customers and share holders. But, this business will have little in common with current software profit margins.

On your second point, you were asking if being "2 years behind is really that bad". There are two components of advantage: 1) technical advantage, and 2) scale advantage. On the former, I agree with you that technical advantage is usually not durable and best measured in time. We could debate the time frame, and there are some advantages where patents and other IP protection apply, but I largely agree with your assessment on technical advantage. On the scale advantage, its huge and no number of years will change that. At scale, telecom, server hardware, networking gear, infrastructure, etc. are just fundamentally cheaper and these are the dominant costs when running at scale. 24x7 security, admin, and support staffs are hard to afford at low scale and irrelevant costs at high scale. These advantages of scale are durable and will always be advantages of public clouds and are the primary reasons why private clouds are not the right approach.

Finally, you ask if a company can do the first move, the move to the cloud, how much harder is it to move to a private cloud and then move again to a public cloud? Essentially, moves are hard. They are always harder than predicted on the way in. They can be done but it will be expensive. I'm arguing that the cost of a move is worth avoiding and the time given up to competitors than went directly to the cloud is my biggest concern.

Thanks for the thought provoking set of issues.

jrh@mvdirona.com

James Hamilton

Comments are closed.