Practical Analysis: Don't Rush To Cloud Computing

Smart organizations will take a measured approach to this latest and greatest IT ''paradigm.'
What's more fun than being called out to a public debate? My colleague Bob Evans and I often see the challenges faced by IT organizations very differently. Bob's a visionary who has spent nearly three decades watching the evolution of IT and imagining how it could be done better, and I respect that. I spent the first 15 years of my professional career in IT, and tend to take a more conservative view on how IT should evolve, and he seems to respect that. Yet I wasn't surprised when Bob asked if he could quote my concerns about the cloud computing phenomenon in his recent column on the subject, where he made some astute counterpoints but also included this line: "The status quo that Art would have those executives cling to for at least a while longer is simply not up to the task." Now there's a public glove to the face if ever I've had one!

That's the sort of false choice a hyped concept conjures. You're either all in for cloud computing right now, or you're clinging to the last vestiges of a bygone era. Ya don't want to be the last one with a horse and buggy, do ya?

Here's the bottom line: Don't worry if "cloud" is low on your priority list. You're in good company. InformationWeek Analytics asked 370 of you to rank factors impacting data center operations for 2010. At the top of the list: constrained budgets, storage growth, and server virtualization. Dead last: cloud computing. In that same survey we asked about moving to public and private clouds. Only 9% responded that they were either using or considering public cloud services, while 19% were using or considering private clouds. Thirty-nine percent said that public clouds would never be used by their organization at all, and 24% said they wouldn't consider private clouds. I'd never say never, but I think this represents healthy skepticism and a pretty typical IT evolutionary process. Evolution takes time.

The first and biggest problem with cloud computing is that it's an ill-defined term. There are the "as a service" definitions, which include software, infrastructure, and platform. Each of these has a very different business value proposition. Sweeping these concepts together as "cloud computing" leaves us with more questions than answers. Then there's the public, private, and hybrid cloud definitions, each of which implies something different, too. Because of the variety of definitions, it's impossible to have a serious conversation about your monolithic cloud computing strategy. So let's avoid the term and look at some of the genres I've mentioned.

Software as a service is well understood and, if the good folks at Gartner are to be believed, it's a $9 billion yearly business. That's not a huge number, but it shows that the concept is valid, and the smaller and newer the customer firm, the more SaaS makes sense. More mature organizations will likely have systems and processes in place that a SaaS application must displace, and therefore the range of questions that must be answered before deciding to go the SaaS route is larger for them. In my mind, there's no right answer when it comes to replacing in-house apps with SaaS, no more than there's a right answer to buying or leasing a car. You have to go through the exercise of evaluating your needs and understanding the risks associated with each option. (If someone says there is a right answer, you know who you're dealing with--think car salesman.)

Platform as a service is becoming well understood, thanks in great part to Apple. In the B2B world, perhaps the best known PaaS is Salesforce's When you go there, you'll find a bunch of interesting applications, but the relationship between them isn't crystal clear. The business model for PaaS becomes much clearer (or at least it does to me) when you look at the iPhone. The Apple iPhone offers application developers access to motion sensors for games, GPS to locate itself in the world, a touch screen, 2G/3G and Wi-Fi wireless to connect to the Internet, and a great development environment for a portable device that even makes phone calls. Apple created the killer packaging for the platform, and developing for it will be strictly an iPhone game. The same is true for If there's no high affinity with the things that Salesforce does, then using its platform for your applications doesn't make sense. That's the piece that was missing from interesting early PaaS offerings like Coghead. All it had was a development platform without the cool ancillary services and huge customer base. Unless as a business you've hitched your wagon to a large partner like Apple or Salesforce, PaaS probably isn't for you.

Infrastructure as a service is one of those things that sounds great but can quickly lose its attraction as you investigate. There's probably not a CFO in the country who hasn't drooled at the notion of dropping data centers off the company's cost portfolio. Data centers are one of those big fat expenditures that often lead to contempt for IT. The well-meaning CFO will probably do the math at some point and ask the CIO a question like, "If Rackspace can lease us a server for $10 per month, how is it that ours seem to be costing orders of magnitude more?" And while you can start the conversation by pointing out that you can buy a terabyte external drive for your laptop for under $100 but the same enterprise-quality terabyte from EMC costs thousands, there's obviously some sense to looking at using outsourced infrastructure for some applications. Again, in the general case, small and startup companies will have an easier time building their businesses around these systems than large enterprises.

The auto buy-or-lease analogy is instructive here, too. Zipcar brought a great twist to the equation by offering cars for $7 an hour. Suddenly, city dwellers who couldn't justify the cost and hassle of owning or leasing a car had an alternative. But the alternative makes sense only for a certain set of would-be drivers. The same is true of IaaS. If you have an application that runs for a long time and takes a lot of CPU cycles but operates on comparatively small data sets, then IaaS may be perfect for you. Need to run a bunch of Monte Carlo simulations once a month on data sets of a few gigabytes? Then step right up. But if you need to run the jobs much more frequently and if you need to operate on multiterabyte-size data sets--and thus, simply transmitting the data to the IaaS provider becomes an expensive proposition--then there's probably a better, more cost-effective solution. These sorts of hit-and-run computing applications certainly exist, but they aren't the norm.

If you simply want someone else to run your data warehouse or other application, you'll find yourself either looking at SaaS or a classic outsourcing agreement, where you specify SLAs and negotiate how your data will be treated while in the outsourcer's hands. While cloud computing subsumes all of these services, it's just as likely that cloud adherents are speaking more conceptually, endorsing the idea of decoupling applications and operating systems from the underlying hardware. Here our survey speaks loud and clear: IT professionals think of this as virtualization, and it's among their highest priorities. If you want to call it cloud, then have at it. But why would you? While it's possible to take advantage of external servers and storage systems without virtualization, it makes a lot more sense to have broken that application-to-hardware bond internally and then go thinking about using compatible external systems.

As useful and important as virtualization is, most IT shops are still in the process of embracing it. Disaster recovery is one of the more obvious places to use virtualization, and yet in a recent InformationWeek Analytics survey, 34% said they wouldn't use virtualization at all, and an additional 21% said they'd use it on no more than 20% of their systems. The No. 1 reason for not using virtualization for DR was that software vendors wouldn't support it. While I don't recommend rolling over for software vendors' outmoded requirements, the stat does show that, yes, evolution in IT takes time.

There will be no "year of cloud computing," not this or any other year. More likely, there will be a decade or so of hashing out what works and what doesn't. Few IT professionals cling to the status quo; those with that personality are usually more comfortable in other jobs. But it doesn't take too many years in the IT trenches to realize that innovation for innovation's sake won't win you many friends. Advising steady evolution of systems and processes done in close conjunction with your business partners will never light up the blogosphere the way a clarion call to the next cool thing will, but that doesn't mean it isn't good advice.

Art Wittmann is director of InformationWeek Analytics, a portfolio of decision-support tools and analyst reports. You can write to him at [email protected]. Register to see all reports at