[ Read the whole CIO.com series, "The Case Against Cloud Computing," by Bernard Golden. See Defining Private Clouds, Part One, Defining Private Clouds, Part Two, The Case For Private Clouds and The Case Against Private Clouds. ]
As you probably know, the conventional wisdom in the technology industry is that large enterprises are going to gravitate toward private clouds. The three reasons most often advanced for this are:
1. Building a private cloud enables IT organizations to leverage existing infrastructure, thereby making cost-effective use of previous investment.
2. Placing cloud computing inside the data center obviates many (if not all) the issues that accompany public clouds, e.g., data privacy.
3. Private clouds can obtain a lower TCO, since they don't have a profit margin added to the base cost structure of a cloud environment.
Less often advanced, but underlying these reasons, is an unspoken belief that large enterprises just won't be comfortable relying on external compute resources and will instinctively trend toward placing a cloud internally.
Based on the assumption that large enterprises will go the private cloud route, most of the major technology vendors have launched private cloud initiatives. For example, just this week IBM announced a service to help companies implement a cloud-based test/dev environment inside the firewall. Its competitors such as HP also have private cloud offerings. One might go so far as to say that the conventional wisdom on this subject is along the lines of "sure, the public cloud providers got this trend started, but now the big leaguers have shown up to do it right—inside the data center."
Therefore, I found the Forrester report "Conventional Wisdom is Wrong About Cloud IaaS" to be compelling reading.
Forrester's Key Findings
Of the enterprises responding to the Forrester survey, about one quarter of enterprises plan to spend or are spending on IaaS via an external service provider. To fall into this group, the enterprise has to be pretty far advanced in its implementation plans—this is not companies in a "discovery" or "evaluation" phase. This is companies who have made a decision to move forward with an external cloud provider, or are in fact implementing a system hosted by an external cloud provider.
Firms are slightly less interested in internal clouds than in external IaaS. By a margin of 10 percent, companies of all sizes prefer to focus on external providers rather than implementing a cloud internally. This is really surprising, for two reasons: (1) it indicates that companies feel they have enough information to make a decision, which is somewhat surprising given how early in the process we are; and (2) despite how early in the process it is, most companies are not opting for the "safer" choice, which is creating an internal cloud.
Firms are interested in an internal cloud or an external cloud but not both. When the percentages of those companies who have selected either internal or external clouds are summed, less than half of all companies would like a mixed cloud environment. Again, this is pretty surprising, since many in the industry characterize a mixed internal/external cloud topology (aka, "hybrid" or "cloudburst") as the most desirable.
Larger firms are more interested than smaller firms in leveraging external IaaS capability. This flies in the face of conventional wisdom that the SMB market will be the most eager adopter of cloud computing, because it will enable them to avoid extensive internal IT investment and skills.
In actuality, it's probably not surprising that this is the case; IaaS requires technical skill to pull off successfully, and SMBs aren't usually that strong in this department. I'd believe that SMBs would adopt SaaS more adroitly than large enterprises, though. After all, by using SaaS applications, SMBs bypass all the technical complexity that accompanies managing servers, whether installed locally or remotely.
Interest in production app placement in external clouds is nearly as high as for test/dev. Again, the conventional wisdom is that companies will migrate test/dev to clouds as the initial use profile, because test/dev is often a pinch point in provisioning, requiring resources quickly and for indeterminate durations; the thinking goes that this type of use profile meshes well with cloud computing characteristics but also sidesteps other issues associated with external clouds like security, data privacy, and SLA needs.
For example, in last week's cloud announcement IBM cited test/dev as the initial application type to move to the cloud, citing this use as a likely first step in cloud use. Instead, this survey found that companies are nearly as likely to put production systems into the cloud.
Assuming the survey is a fair reflection of end user sentiment, this may be one of the rare cases where end users are actually ahead of the vendor community. Many vendors maintain that businesses will take a measured -- if not reluctant -- approach to cloud computing, with the path to the cloud beginning with low-risk applications like test/dev hosted internally. This survey pool seems ready to move forward much more aggressively than the common wisdom expects.
Lessons to Take Away From This Data
There are a number of implications one can draw from this report.
First, something is driving these users toward cloud computing. Across all sizes of companies, the percentage of "not interested in cloud computing" ran about one-third; astonishingly small for such a new technology initiative like the cloud. Clearly, strong dissatisfaction with the existing method of provisioning and operating compute resources is evident.
For many IT organizations, the cost and complexity of IT infrastructure is outstripping the ability of the organization to manage. There's a mismatch between traditional methods and future demand; it will be extremely uncomfortable for any IT organization hapless enough to get caught between those two imperatives.
Second, predicting what users will want is troublesome. Many cloud vendors assert that they're going to help IT organizations build internal clouds, with a confident assumption that companies will want to leverage installed infrastructure and make it more agile. This overlooks the fact that most IT organizations don't want to disrupt working systems, and grafting cloud capabilities onto an existing infrastructure will inevitably cause disruption. Just because vendors want customers to upgrade everything doesn't mean that users want to do so.
Third, end users seem much less conservative in their technology plans than one might expect. Instead of the test/dev baby step, IT users appear to be ready to move forward with production systems. This means that for them, the general question about cloud computing is already answered; what is left is technical and tactical issues—the how, not the what.
Even I was surprised to read the stats on this. Most surprising: more than one-third of both large and medium enterprise companies are ready to put enterprise applications into production in external cloud providers, according to the Forrester survey. Read that again: more than one-third of both large and medium enterprise companies are ready to put enterprise applications into production in external cloud providers.
I can't tell you how many conversations I've had with cloud skeptics pronouncing that "companies aren't ready to put enterprise apps into the cloud"—some even go further and proclaim "companies won't put enterprise apps into the cloud" with a finality that brooks no disagreement. I guess, based on this survey, that 19th Century American humorist Artemus Ward's observation is only too accurate: "It ain't so much the things we don't know that get us in trouble. It's the things we know that ain't so."
Overall, this survey aligns with my experience with cloud computing. Cloud computing moves from "unheard-of" to "must-do" much faster for IT professionals than I've seen with previous platform shifts. I attribute this, as I said earlier, to a widespread dissatisfaction with existing IT practices. The demand for IT services is outstripping the ability of IT organizations to manage using traditional practices.
Something's got to give, and it seems that Forrester has uncovered evidence illustrating that fact.
This is fine if you want to add lots of front end systems doing the same thing. But in a traditional two tier architecture, putting your databases on the same network as your front end web traffic creates all sorts of headaches. First of all, while you can secure the servers it's generally best not to directly connect sensitive database servers to the internet.