Blog Post

Is the Cloud Right for You? Ask Yourself These 5 Questions

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Is cloud computing right for you? For the fledgling startup, the appeal of the cloud is obvious. Given how easily an entrepreneur’s vision can be stymied by a lack of technical and operations expertise, leveraging an Amazon EC2 or Google App Engine could provide the only viable option.

But what about large enterprises that not only have an in-house technical staff to do their bidding, but existing data centers and deep pockets? Stacey has already identified issues with some cloud providers, such as security, reliability and portability. However, assuming they are all resolved, are there compelling reasons for large enterprises to even be interested in cloud services? And if so, under what conditions?

In order to decide, the enterprise needs to ask these five simple questions:

1. Is demand constant? If demand is constant, dedicated resources in an enterprise data center are fine. Smooth, constant workloads mean that a fixed pool of servers can chug away 24/7, meeting utilization targets.

Unfortunately, very few enterprise applications have this kind of profile. Consider a retailer that does 80 percent of its annual business in the month following Thanksgiving Day. Fixed capacity engineered to peak would only be fully utilized during those four weeks, compared to utilization of slightly more than 2 percent during each of the other 11 months. In other words, its investment would go virtually untouched for more than 90 percent of the year. Try selling that to the finance committee. A cloud, on the other hand, can provide dramatic cost savings by offering access to scalable resources on a pay-per-use basis.

2. Is growth predictable? Even if demand isn’t constant, if growth is predictable, it can be managed in an enterprise data center. By building in lead times, one can place orders for additional capacity and rest easy that it will arrive in time, even by snail-mailing paper purchase orders for servers that get delivered by slow ships.

But when growth is unpredictable, the pay-per-use model of the cloud makes “cloud-bursting” — that is, leveraging cloud services to handle spikes — a more cost-effective option. Plus, with fixed capacity, it’s easy to make one of two fatal errors: Being overwhelmed by surprise demand can easily lead to poor performance or no service, resulting in the loss of both revenue and reputation. On the other hand, investing in capacity that remains idle can bankrupt a small company.

3. Can demand be shaped? Users today expect instant gratification, even from free services. While some demand can be shaped and smoothed – either by avoiding it, deferring it, or incenting it, via sales, promotions, queuing, congestion pricing, or variable pricing for yield management — some can’t.

If spiky demand can’t be shaped, on-demand scalability is indispensable. After all, how popular would Google be if a search request returned, “We’re kinda busy right now. How does next Tuesday around 2 pm work for you?”

4. Where are the users? If users are concentrated in a particular locale, they can be serviced by a single nearby data center (or two, for business continuity), but not if they’re scattered around the globe.

The only way to engineer today’s rich Internet application for a global community is to leverage a network of dispersed web, content and application servers. While building lots of data centers all over the world might have seemed like a good idea for enterprises a few years ago, a better option today is to consolidate enterprise data centers while simultaneously leveraging a cloud service provider with a global footprint.

5. Is the application interactive? There are still many applications out there that aren’t highly interactive, such as seismic analysis, circuit simulation and equity portfolio optimization. For them, geographic dispersion could be a negative, due to an inability to effectively partition the compute tasks into loosely coupled components.

However, for the tidal wave of emerging Web 2.0, AJAX, Rich Internet Applications, proximity to a service through geographically dispersed cloud resources is key as the client portion of the app has to make frequent round-trips to the server — in some cases on every keystroke. And for applications such as multipoint video collaboration, reducing hops and propagation delays between end points and cloud-based bridges is essential to creating a natural experience.

So for enterprises with smooth and predictable demand created by accommodating users who are willing to walk back across the street another day to process their batch jobs, clouds may be unnecessary.

But for enterprises pursuing emerging, shifting and uncertain global markets, with global supply chains or virtual enterprise partners and variable and unpredictable workloads coming from demanding users who want engaging, interactive interfaces, the cloud could be the right — perhaps even the best — option.

Key Question Enterprise Data Center Better Cloud Services Better Key Cloud Benefit
1) Demand Constant Variable Scalable and On-Demand
2) Growth Predictable Unpredictable
3) Fungibility of Demand Deferrable or Promotable Not Shapeable
4) Users Concentrated Dispersed Globally Dispersed to Reduce Latency
5) Applications Batch Highly Interactive

Joe Weinman is Strategic Solutions Sales VP for AT&T Global Business Services. The views expressed herein are his own and do not necessarily reflect the views of AT&T.

19 Responses to “Is the Cloud Right for You? Ask Yourself These 5 Questions”

  1. Bob Johnson

    I’m glad you restated the focus of the article in your comments – most people seem to agree that on-demand services make sense for small companies, but whether they do for large enterprises is very much up in the air. I’ve been promoting a plug-in,, but was intrigued by the idea, put forth by one of your readers, of QA as a service. I can see it working well for small companies, but I don’t really see it as an option for mid-size companies, let alone, large organizations. Looking at your questions, demand for QA may not be constant, but it is predictable and can be shaped (planned for), so in-house resources may not be the answer, but out-sourcing or hiring contractors does make sense, and probably more sense than on-demand services.
    On-demand, sales automation, on the other hand, looks like it is poised to make inroads into mid-sized and larger organizations. In another life, I worked in a very large retail organization, and what I saw over a period of years were cycles of “clump it” (centralize) then “break it” (disperse), then repeat. This is common to many industries. The nice thing about on-demand software, it’s adaptable to both cycles. SalesForce .com can provide an organizing principle for sales departments that are trying to centralize – without creating a burdensome infrastructure. And the plug-in model works very well for dispersed organizations. One sales group may want pipeline management whereas another group wants to automate lead generation, . The Cloud solution called SalesForce supports both for big, medium and small companies.

  2. Alistair: I think we are completely in agreement. If I can summarize your response, it is that I am correct for large enterprises, but that there are additional advantages driving smaller companies to the cloud.
    But if you review my introductory paragraphs, the focus of my article,
    and the exact question I posed, was:

    “are there compelling reasons for LARGE enterprises to even be interested in cloud services? And if so, under what conditions?”

    A large enterprise has almost certainly adopted virtualization technologies: as an example, VMware reports that they have 100% of the
    Fortune 100 as customers, and 95% of the Fortune 1000 (The remaining 5% are probably also leveraging virtualization, they just may not
    be VMware customers). Consequently, a cloud service provider does not generate economic value for such a customer merely by enabling

    Large enterprises also typically have extensive technical capabilities; the ability to exploit volume purchasing agreements; the ability to leverage economies of scale and operations automation; and standardized development frameworks, environments, and tools. Consequently, I don’t believe that those attributes in and of themselves are compelling for most large enterprises to pursue a cloud-based strategy.

    The executive summary of my original article is: smaller companies, including startups, clearly benefit from clouds for the reasons that you and I mention; AND, large enterprises — even with substantial deployments of virtualization, deep technical capabilities, volume purchasing, standardized environments, data center automation, and everything else — can ALSO benefit, primarily by leveraging the cloud’s scalability coupled with a usage-sensitive pricing model for bursty, unpredictable, and/or uncontrollable demand, and by leveraging a cloud’s geographic dispersion, especially for highly interactive applications accessed by a global user base. There are a number of such demand types: seasonal spikes, quarterly book closes, special promotions and events, and ones that you mentioned, such as QA. Lastly, the vast majority of the largest enterprises are already primarily global, or are increasing their global presence, implying globally dispersed users, such as employees, supply chain partners, customers, or other stakeholders.

  3. I agree with the state of “QA is a gateway drug for cloud computing”. Check out for a good example of how to use the cloud to improve your product quality. It’s a known fact that today’s QA challenges are almost impossible to cope with. Having the back of “the cloud” can be a great extension to every company testing activities.

  4. @Joe: Interesting piece, but I think it overlooks a number of other reasons why clouds are catching on with mid-sized companies. What you say is true largely for the Global 2000, not the world at large.

    First, I think experimentation is a big factor in cloud computing. IT is in a constant struggle between efficient use of capacity (have a few machines and maximize them) and ability to respond to new service requests (have spare machines ready to go at the risk of being idle.)

    The comparison you’re making is really one of a truly virtualized private data center (with CPUs running at 80% or so) versus a cloud environment (pay as you go.) Prior to virtualization, it didn’t make sense to run your own data center. So I would offer that full virtualization is a precursor to your arguments, and many smaller companies aren’t virtualizing their compute infrastructure. For them, moving to a cloud environment gives the economic benefits of virtualization.

    Second, we need to consider operational effectiveness. If a company is big enough to hire specialists, fine. a huge investment bank may have several full-time developers building custom operational software, or the leverage to beat up suppliers. But for most firms, they can’t compete with the buying power or economies of scale of a cloud’s operations.

    It’s not just about computing utilization; it’s about overall efficiency. If something is a commodity — as computing is today — then it needs to go where it’s most efficient. If that happens to be a damn in Les Dalles, that’s the right place for it.

    Third, we have to look at the ability to develop quickly. Clouds aren’t just virtual machines. They’re other services, such as storage (Bigtable) and message queueing (SimpleDB.) If that accelerates development and reduces the need for internal employees to run things (such as doing database administration) it will often be a better choice.

    And fourth, there are parts of an IT organization — specifically, QA, testing, and training — that will never have constant demand. They are by nature bursty. It’s one of the reasons Skytap is getting so much traction: They focus on the bursty QA department. QA is a gateway drug for cloud computing.

    Your comparison model is a good one, but only for really big companies that have already bought into virtualization and that have a staff as comprehensive as those of a cloud operator.

  5. Jack Johnson

    @Joe: Very interesting post. As you clearly mention, cloud computing isn’t for everyone. In some (most?) of today’s cases, they are good enough with their own data center.

    Although I see a trend to more and more applications and services needing to be moved, because of their increasing demand.

    If they need to respond to some relatively new requirement on IT or/and each will require a different sort of control over corporate data that was previously necessary. Some concrete application the author mentions is e-Discovery, as this one is taking some time for the IT & Legal departments alike.

  6. @Joe, that makes sense. If you’re referring to the geo-location + latency + interactivity equation then it makes sense cause the cloud will (depending on the vendor) give you global presence and lower latency. Guess I read your post as just High Interactivity vs Batch.

  7. Excellent post, Joe made more substantial with both comments by Michael and Darayush! Your thoughts would make a nice primer/ overview (along with Stacey’s)for companies to seriously consider these factors before jumping into the cloud bandwagon as opposed to just following merely because of the hype surrounding it.


  8. I’ll address Darayush and Michael’s comments together. My focus in this post was not on the benefits of clouds, but rather to broadly characterize the conditions under which clouds can generate those benefits. Enterprises are typically interested in cost reduction, and also revenue growth, margin enhancement, risk reduction, customer loyalty, employee satisfaction, regulatory compliance, and corporate / community responsibility. Cost reduction, which I did address in the post, does not arise merely from virtualization or shifting computing, storage or applications to the cloud. Enterprises are very adept these days at exploiting virtualization to enable server or storage consolidation, and therefore attain enhanced utilization as well as cost and carbon footprint reduction. For enterprises, clouds drive additional benefits in this arena through economies of scale, and also via the statistics of scale: statistically multiplexing uncorrelated demand from multiple companies, applications, or users, thereby enhancing utilization, in turn providing a globally cost-optimal solution for both service provider intermediaries as well as the enterprise customer. I provide a detailed analysis of this at my web site and tools for analysis of these effects at my simulation site. Enterprises have access to the same hardware infrastructure, management tools, hypervisors, virtual server management and migration tools, etc., that cloud service providers do. Therefore, the economic and carbon advantage of clouds to enterprises does not arise from those technologies themselves, but rather ultimately derives from demand multiplexing and scale economies.

    Daryush, my argument was that highly interactive applications supporting a global user base ultimately require a geographically dispersed architecture. An enterprise could try to invest in creating one, but most enterprises these days are going the other way: consolidating dozens of worldwide data centers down to a few. To meet the latency requirements for round-trip transactions for today’s applications, it is better to “rent,” i.e., leverage the cloud, than to own. This is exactly the same logic as staying in various hotels in different cities (a pay-per-use strategy on dispersed resources) during a round-the-world trip, rather than owning a house in each location you might visit (owned and dispersed), or trying to commute home every night no matter where you spent the day (owned and consolidated but high latency).

  9. Great concise post. Two things that come to mind, one factor that I didnt see mentioned was Cost.
    The other one was the classification that “Highly Interactive” applications are go better with Cloud Services.
    A box is a box and the S/W on it is the S/W on it, services in the cloud are still HW and SW in a data center shared by multiple parties, so I’m not sure how the interactivity of the application works well just cause its in the cloud. Or are we trying to relegate Enterprise Data centers to mainframe by saying they are good only for batch?

  10. Interesting post Joe. You ask some good questions and provide a basic framework for evaluation purposes.

    There are, however a couple more items that I believe you neglected to mention, namely Cost, “Going Green,” Obsolescence and under-utilization of data centers. (There are probably others as well.)

    Corporations are concerned with costs. Data centers are expensive. Outsourcing your data center is costly as well. Creating Cloud Data centers is a cost-effective method for reducing some of your burn.

    Traditional rack and stack infrastructures are not Green. The Green-ness of your corporation will be come increasingly important. Using virtualization of some sort or another (e.g., through Cloud Infrastructures like GoGrid) will reduce your carbon footprint significantly.

    Corporation cannot keep up with the latest and greatest software and hardware. It’s cost prohibitive and a human resource drain. Outsourcing to Cloud providers allows for the use of the latest tech much more easily than what was traditionally.

    Data centers are very underutilized. Corporations should look hard at reducing their CapEx by dumping old systems and downsizing their data centers, or, as I have been saying, outsourcing to the Cloud for many various efficiencies.

    There was recently an interesting article on Forbes about the rising costs of data centers. It too, however, neglected to discuss cloud computing as a viable option for the enterprise. I have broken down points of the article in my blog post here:

    Linked blog post

    Thanks for the article though, I think it will get a lot of people’s attention.

    -Michael Sheehan
    Technology Evangelist for GoGrid (Winner of LinuxWorld 2008 Best of Show)