Thursday, February 11, 2010

From private clouds to solar panels: more control and uniqueness, but are they worth it?

Andi Mann of EMA wrote recently that failures are endemic to public clouds. And, by the way, that’s OK. In fact, says Andi, failures are normal part of what your IT infrastructure needs to be able to deal with these days.

Even if you take it as a given that we’ll hear about cloud service failure after failure in the news from now on (a daunting prospect in and of itself), public clouds surprisingly still set a pretty high bar for internal IT. Andi’s figures put some public cloud uptime numbers at 3 to 3.5 “nines,” as in 99.9 or 99.95% uptime – that's 5-10 minutes of downtime each week.

Now, if you’re hoping to get a lot of the public cloud computing benefits but do so on-premise by creating a private cloud infrastructure, there's a serious amount of work and investment required to match public cloud availability for all of your applications. Andi pins “normal” cloud computing outages at 5-10 times less likely than those in internal data centers.

“CIOs who are planning to build their own private cloud have a surprisingly high bar to reach,” blogs Andi.
Sounds like it may not be worth the effort for private clouds then, eh?
Actually, think again. It just might be, but for other reasons than you might think.
Private clouds hold what’s most unique about your organization
Mike Manos recently had some interesting observations from his time at Microsoft and at his recently-ended stint at Digital Realty Trust (sounds like he’s heading to greener pastures at Nokia). In response to James Hamilton at Amazon (thanks to Dave O’Hara for pointing out the discussion), Mike postulated that the things that make private clouds quite interesting, despite the high bar, are the way they encapsulate the unique components of an organization.

In other words, the most tailored and specific things about your IT environment are the best argument for a private cloud.

That rings true for me. Cloud computing is a way to pay for only what you need, and a way to have compute, storage, and other resources appear and disappear to support your demands, as those demands appear and disappear. There are components of what IT does for your company that are not unique. Those sound perfect to move to external, public cloud computing services at some point. They are commodities, and should be handled as such. Maybe not now, but eventually.

The more specific, complex, special pieces of IT seem logical to be the ones you keep inside as you get started down the cloud path. Those take the kid gloves and your special expertise, at least for now.

The push to get the most from those important, unique pieces of IT is giving enterprises a strong incentive to pursue a cloud-style architecture in-house. To again quote Andi Mann’s EMA research, private clouds are the preferred approach to cloud computing for 75% of the organizations they polled, far ahead of the interest in public clouds.

Are private clouds a temporary fix or a permanent fixture?

With all of that as a background, how permanent are private clouds? Here's a quick detour to help answer that:
Chris Hoff of Cisco collected commentary at his Rational Survivability blog on the topic of private clouds recently by weighing in on the appropriateness of the IT-as-electricity analogy that Nick Carr brought mainstream with The Big Switch. His quick take was that private clouds might be like batteries (thought he didn’t go too far explaining his concept, beyond labeling it an “incomplete thought” to get conversation going). However, a couple of his commenters had an analogy I liked better: that of a solar power generator.

So, is a solar power generator a good analogy for a private cloud? You’re generating “IT power” for your own use, using your own resources. Unlike what the “battery” analogy implies, a private cloud implementation is not what you’d call temporary. In fact, as Manos was thinking about private clouds in the blog noted above, one of his comments was that “there’s no such thing as a temporary data center.” Or a temporary private cloud infrastructure, I’d add. Like it or not, most IT projects, even if they are done for an ostensibly short period of time, end up living long beyond their intended sunset.

Private clouds will be no different. Many (like Gartner) see private clouds as a stepping stone or a temporary requirement until the public cloud addresses all of the roadblocks people keep complaining about. But once that infrastructure to add/subtract, build/collapse things in your IT environment is in place, you should be able to get a lot of use from it. And it will live on. This is similar, in fact, to the situation if you had taken the time and effort to get that solar installation up and running.

As things progress, I predict this “either/or” kind of language (as in “public clouds” or “private clouds”) that we’ve been seeing will fall by the wayside. I think Manos is right: “and” will be the way of the future. We’ll aim for use of the public cloud where it makes sense. And we’ll keep using private clouds – leveraging their reflection of organizational uniqueness, coupled with an unintentional permanence because, well, they work. We’ll find a way to take advantage of both.

Paving the way for hybrid clouds

This scenario, by the way, makes hybrid clouds the end state, a situation that Hoff sees as “the way these things always go.” Scott Hammond, CEO of newScale, uses an example that reiterates this: “The data center looks like my dad’s basement.” In other words, IT continues to be a strange mishmash of the new, combined with all that’s come before. That’s reality.

So a conversation that started by questioning whether public cloud computing service outages are endemic or even a problem, which then shifted to how private clouds can hold unique value for organizations, ends up connecting the two.

Of course, hybrid clouds will require an additional level of thinking, management, and control. That’s a topic that will have to get a unique post of its own one of these days.

In the meantime, I’ll leave you to ponder what other cloud computing metaphors we might be able to unearth in Scott’s dad’s basement. It just might be worth it. Especially if we find something to help those solar panels pay off.

No comments: