6 Half-Truths About the Cloud

30 Comments

Is there any subject in IT today with more promise — or more confusion — than cloud computing?  Here are six commonly held views that, while not wrong, are just not entirely accurate:

1. Economies of scale are the key to cloud benefits

A number of industry analysts have argued that public infrastructure clouds generate value for customers through unit cost efficiencies derived from massive web-scale operations. True, large service providers can achieve economies of scale through volume discounts, experience curve effects, and consolidate operations at a scale that minimizes unit costs.

However, enterprises and cloud providers build data centers out of the same components and similar tooling, purchased at similar price points. While some economies of scale may exist, a greater role may be played by the statistics of scale: By aggregating uncorrelated demand from multiple customers, service providers can reduce variation and maximize utilization. After all, a retailer’s peak after Thanksgiving is different than a tax preparer’s peak in early April.

2. All IT will move to the cloud

Nick Carr, in “The Big Switch,” highlights the parallels between electricity and IT. When electric power was first adopted by industry, companies owned their own generators. This approach rapidly shifted, due to cost-effective AC power generation and distribution technologies and today, of course, almost everybody buys their power from a utility. Conclusion: the same will happen to IT.

True, cloud-based IT services offer a number of advantages.  And no one can argue with results: A number of enterprises have eschewed in-house deployment of, say, CRM or messaging because it’s just not worth the upfront licensing costs, upgrade and patch hassles, or the effort to keep the environment operational.

However, web-scale cloud infrastructure is largely built out of the same servers and storage available to enterprises.  This differs from electricity, where, say, a desktop nuclear power plant or personal hydroelectric dam with roughly equivalent cost per kilowatt-hour does not (yet) exist.  An economic optimum occurs for infrastructure when owned resources are balanced with cloud resources in a proportion that depends on the utility price premium and the peak-to-average ratio of demand.  Consequently, perhaps housing is a better parallel than electricity: It’s best to use a thoughtful mix of your own resources and hotel rooms.

3. Clouds generate value by replacing capital expenditures with operating expenditures

True, avoiding a large capital outlay may make a difference to two teens with their Mom’s credit card and an idea for an innovative Facebook application.  And capital expenditures imply an investment in fixed capacity, whereas cloud services offer the promise of agility.

However, for solvent enterprises, using capital vs. cash flow is a decision that won’t impact the key metric for any well-run firm: Return on Invested Capital. The point is that capital expenditures are not evil contraband and operating expenses aren’t manna from heaven; choosing between them is a financial decision to be addressed by the CFO.

4. Private clouds are as effective as public clouds

True, a “private cloud” can use dynamic resource allocation, virtualization, chargeback mechanisms, autonomic provisioning and performance tuning, just like a public cloud. And sharing resources across business units is certainly better than siloing them. Also, the ability to seamlessly cross public/private cloud boundaries based on real-time demand can create real business value via cloudbursting.

However, with a “private cloud,” an enterprise still must expend capital in fixed quantities (or, equivalently, commit to fixed lease payments) regardless of utilization, removing the key benefit of nearly unbounded scalability without prior investment or long-term commitment, a key mechanism of value generation by service provider clouds. I’ve defined a cloud to be a CLOUD: a Common Location-independent Online Utility service, available on-Demand.  Not only do some IT and telecom services fit this definition, so do electric utilities, rental car chains, and hotel franchises.  By analogy then, a “private cloud” makes as much sense as would be something like a “personal hotel.”  This is more than semantics — it’s economics.

5. Cloud = virtualization

True, cloud services benefit from virtualization, and hybrid architectures that seamlessly integrate in-house capacity with cloud services will enable cloudbursting to reduce cost while enhancing flexibility.

However, virtualization is orthogonal to cloud implementation. Enterprises can leverage server and storage virtualization in a fully owned enterprise environment, or cloud service providers can leverage virtualization in a shared services environment. The use of virtualization, then, is unrelated to whether the environment is a cloud or not.

6. Clouds are greener

True, they may reduce total power consumption and therefore carbon footprint.

However, this is only because multi-tenancy and aggregation of uncorrelated demand go one step beyond virtualization, reducing total resource requirements through statistical multiplexing. Fewer resources imply less manufacturing, and less electricity consumption for power and cooling.

Joe Weinman is Strategy and Business Development V-P for AT&T Business Solutions.

30 Comments

judy

if I use DB platform that stores data in their cloud and charges users /month cost – is this potentially more expensive to the user than if I use a different type of DB that is hosted with my website?
Judy

Gunga

“However, web-scale cloud infrastructure is largely built out of the same servers and storage available to enterprises.”

Well, you are missing the point entirely : they aren’t. If you look at Amazon and Google, they rely on their own technologies. In Google’s case we now know that hey don’t stop at software to accomplish this, but that they build their own datacenter technologies — right down to the motherboards on their custom-designed servers. There is precious little in a Google datacenter that you and I can buy today.

Pay attention in class.

You could not get further from “enterprise software” if you tried and implying that what powers the cloud (for a reasonable and practical definition of the term rather than the fantasyland one) is the same software that powers the enterprise today embarrasingly reveals that you know precious little about technologies that have already been deployed.

The closest thing to enterprise components being deployed is Microsoft’s cloud offering — although that is mostly talk for now. Nobody really knows how Azure will stack up against the competition and the verdict won’t be in for another couple of years. There are some aspects of Azure that are exciting and probably the right thing to do; such as their chosen level of containment — but currently you do not find this model in the enterprise in any significant way.

Treova

Nicely written Joe – these assumptions are not necessarily inaccurate either. It all depends on your perspective of cloud computing:

1. Multi-tenancy is key to cloud optimization from a public cloud-provider’s perspective. Notice the cost differences between dedicated servers versus shared servers facilities from ISPs. If multi-tenancy is not an option, then public cloud computing gets reduced to a mere outsourcing of facilities and operations.

2. Overtime, most IT will move to the cloud. It is self-serving to assume that IT runs a business…it is the other way around…IT should support business initiatives. Bulk IT services that are needed by everyone – desktop management, email management to run of the mill E-commerce, etc can (and should) be procured through public clouds. Business capital should be expended for IT enhancements that can truly provide a strategic advantage.

3. For most businesses, cloud computing is likely to be less expensive than owning the entire IT infrastructure. However, for companies that are very large, this is not likely to be the case. These largely companies are likely to acquire IT capabilities in bulk or have in-house capabilities because of higher value provided.

Your points are true for many cloud subscribers….it all depends upon their characteristics.

Disclaimer – Treova (http://www.treova.com) is a cloud computing VAR.

Ed

The comparison that “highlights the parallels between electricity and IT” may not be one to write in stone since there is a push for more distributed generation – combined heat power, fuel cells, photovoltaic systems and concentrating solar power, microturbines and turbines, biomass, wind power, etc.

According to the DOE:
“Distributed power generators—small, modular electricity generators sited close to customer loads—offer advantages that large-scale, capital-intensive, central-station power plants cannot provide.
By siting smaller, more fuel-flexible systems near energy consumers, distributed generation avoids transmission and distribution power losses and provides a choice of energy systems to the utility customer. Many distributed power systems produce so little noise or emissions that they can be located inside, or immediately adjacent to, the buildings where the power is needed. This greatly simplifies the problems of bringing power to expanding commercial, residential, and industrial areas.
Distributed energy systems offer reliability for U.S. businesses and consumers who need dependable, high-quality power to run sensitive digital equipment and can provide alternative, less-expensive power sources during peak price periods. The potential market for providing power during peak price periods is as high as 460 GW, according to a DOE study.
Distributed power generation technologies use a variety of fuels, including natural gas, diesel, biomass-derived fuels, fuel oil, propane, hydrogen, sunlight, and wind.”

rohit

Good article.

@Ophirk, I see the network as an integral portion of any cloud offering. Status quo on WAN is not enough. Latency can be managed if it is deterministic. WAN as dumb pipes won’t work – a smarter network is required. Amazon, Google, Microsoft/Yahoo, Akamai, and perhaps even SUN (should it get its act together or hawked) need to pay attention to the network side sooner rather than later. I believe rapid development on the server/app side has left networking vendors and operators a generation behind.

rohit

Pete L

I couldn’t disagree more strongly with your first point. Economies of scale are the primary driver of cloud computing today. Most cloud computing clients are smaller enterprises, not fortune 500. This scale comes into play in many many ways.

1) Hardware. Even given the same applications, cloud providers have the scale to make hardware like SANs cost effective. Certainly you pay more per GB even at scale, but if availability is important to you and you want to reduce your backup costs, etc…. this starts to become a must for critical applications (like email, accounting packages, etc).

2) Labor. This is almost certainly the biggest cost for most small businesses. A cloud provider can afford to employ specialists and well utilize them (even keep them available around the clock). Smaller operations have to rely often on just one or several people whose knowledge is much more generalized and their attention is divided amongst many other competing responsibilities (which is not always efficient). Likewise, they can afford to staff cheaper labor to handle the mundane, train them, setup systems to facilitate it, etc. Even assuming the same skill level and hourly rate, the cloud provider can automate or at least reduce to a science many tasks that would take a smaller business far far longer.

3) Data center costs. Cloud providers can simply do this more cost effectively due to their scale. Consider the cost of getting a proper backup generator, testing/maintaining it, fire control, security, etc.

4) Design of software. Large cloud providers can afford to write their own software from scratch or at least modify theirs which allows them to do things like run on true commodity hardware instead of far more expensive server hardware. This can have a HUGE impact on hardware costs alone, not to mention create greater availability.

5) Internet connectivity. Cloud providers can purchase more bandwidth more cheaply and do so with far greater redundancy.

6) Uncorrelated demand efficiency… is really an economy of scale by another word.

In any event, I don’t personally believe that cloud computing will takeover ALL of IT any time soon. There are some tasks that are too specific and too critical to the company in question to be done well by a cloud provider (as it’s envisioned today). That being said, there are still many many servers being delivered poorly in-house that are completely standard and could be done better and more cheaply by a cloud provider (if said company takes an honest look at their real costs).

George Reese

CapEx vs. OpEx matters, even if you are a big company.

In this economy, credit is hard to come by. Even if you are flush with cash, ROI for CapEx includes the cost of capital (what you could have done with that capital had you not wasted it on a bunch of servers).

Finally, don’t underestimate the value to the CFO of understanding that the cloud servers they are spending money on are actually justified by operations versus being shiny new toys with extra blinking lights.

ML

I usually enjoy your articles. The last one about economic principles of cloud computing was insightful. I don’t quite understand the point of this one though. It seems to me #1 and #6 are at least whole truths based on your explanations. On #6 you basically gave a reason as to why Clouds Are Greener. On #1, cost reduction due to resource pooling is part of economies of scale, isn’t it? You have lower unit cost due to your larger scale.

On #3, I must disagree there. I think that might be the biggest value. You wrote: “However, for solvent enterprises, using capital vs. cash flow is a decision that won’t impact the key metric for any well-run firm: Return on Invested Capital.” Why so? If you have lower unit cost (due to “statistics of scale”) and lower risk (from avoiding over or underspending on your capital investment or from making bad capital investment), you should improve your Return on Invested Capital, should you not?

While I don’t get this one, I am looking forward to your next article!

Personal computer

Cloud computing can be the future with big benefits (as you have listed) but there are still big questions open like security. I don´t think that companies want to loose secret data to their competitors! Time will show us if cloud computing is only a hype or the next generation.

Peter

Thank you for making some great points! I think there is just a major disconnect between the sales side and the IT side of things when it comes to cloud computing. Everyone is so eager to try the cloud, and cloud companies don’t want to leave money on the table, so we are seeing lots of promises that just can’t be kept. I’m not sure I agree about the private cloud part – this new way of provisioning cloud infrastructure – if done right – does address some of the security and privacy concerns around the cloud:

IT can’t see through the cloud

I recently found a private cloud that looks like a reasonable balance of scalability and security. Read this article on persistence in the cloud at the Mosso blog for more on their approach to cloud computing.

Brad Johnson

Great, pragmatic viewpoint. As a cloud customer leveraging multiple platforms to deliver cloud testing services, and thus seeing varying levels of optimism, adoption, trepidation, and rejection of cloud from all size customers, laying out the basic truths of the state of cloud computing today in the fashion Joe has is refreshing. There is no IT nirvana, never has been — just a seeking of a balanced state in between.

Emil Sayegh

Good job in debunking some of the misconceptions about Cloud Computing. This will help in getting all of us on the same page.

On 1, that is a good clarification. Economies of scale no doubt exits with large providers, but statistics of scale play a huge role as well when it comes to hardware. I covered some of this in blog post in February (http://blog.mosso.com/2009/02/cloud-economics/). Your distinction, thought, make perfect sense.

The economies of scale really start to materialize when you start to add up the human resources needed to maintain your own infrastructure. A service provider can leverage, and get much deeper/borader expertise than a single enterprise can in all the different technologies. Especially, if you want to have mission critical apps, and want 24x7x365 coverage for the different technologies.

On 2, I think that Hybrid Hosting, with a mix of cloud AND dedicated (whether your own gear on premise or hosted off premise) will be the norm for larger configurations who have more complex requirements. However, I do think that SMBs will move much faster totally to the cloud than enterprises due to the fact that they many not have as many complex/custom requirements. In reality, anyone developing an app now, wether a startup or large corporation is better served to think about starting it in the Cloud.

Good discussion. Keep up the good work.

Emil Sayegh
General Manager, Mosso | The Rackspace Cloud
emil.sayegh@rackspace.com
http://www.rackspacecloud.com

Prasanth

You “may not” save much on hardware (still need more details to prove it), but, you will definitely save on the army of people needed to maintain your own infrastructure.

ophirk

Great Article.
Few more:

* Clouds are less secure than on premise solutions –
This is partially true as the infrastructure clouds today don’t offer strong authentication and encryption solutions.
However, SaaS solutions like SalesForce.Com demonstrate a very well security architecture that is probably stronger than most on premise solutions that are made available over the web.

* Clouds can not run enterprise applications without complete rewrite –
This generally true for Amazon,Google and Force.Com however this where virtulaization and cloud make a difference and are not just orthogonal. Virtualization provides the abstraction that makes moving apps to the clouds quite seamless with services like IT Structures provide ( http://www.itstructures.com ).

* Clouds enable the of use best of breed solutions –
One nice element of cloud is the fact they are on the web. As trivial as it sounds it enables not just economies of scale but also best of breed solution and free competition.
For example , if one wants to embed document management solution it is now much easier to choose an PAI enabled solution like google docs.Because of the big demand curve there is a place for small vendors that develop simple solutions that otherwise would not be economical.
PAAS solution on the other side enable setting them up with very small investment.

* Latency is not dead ( yet ?)
In the power world solar panels are mostly useful when they installed on premise.
This is because moving power around isn’t that simple.
Similarly there are tons of enterprise applications which are still WAN challenged. Despite all the WAN accelerators, it is not trivial to move legacy applications to the cloud.

Alan Wilensky

Ya, Ya, Ya….

I just performed the cost workups for 2 clients to move dedicated servers inhouse, to Gogrid. When the licensing and projected run time is tallied up, it is not a pretty picture. We are not there yet cost wise. The testimonials at most of the cloud providers web sites are all non-transactional, non-critical, non capital line of business applications.

More research to come in a post on this subject.

Stephen

Alan,

Clouds may still be prices too high for ERP applications, however this could be more cost effective now for office applications, like word processing, spreadsheets…. Finally, no more upgrades of MS Offices every two years for 60 workstations!!!! Taking the money out of Bill’s pocket and putting it in ours!…

Comments are closed.