In the CIO.com article “What Economists Can Teach Us About Cloud Computing”, Bernard Golden paints a fairly well-reasoned picture about the potential future for public cloud versus internally owned IT infrastructure. While I agree with many points in his article, I feel there are a few minor but still critical considerations that need to be made.
The age old argument of Public Cloud vs the other stuff
Bernard and I have had many cloud conversations over the years, but there was one in particular that occurred back in 2011 at a coffee shop in San Carlos California that I think is most germane to the above article. Specifically we talked about whether public cloud would kill any notion of the potential need for or the benefit of private cloud. Bernard felt at the time that public cloud demonstrated an ease of accessibility and use that far exceeded anything available to internal IT and as such would gain and keep the upper hand and effectively kill private cloud before it could even start. Private cloud such as it was is too hard to build and too far behind in its feature set compared to public cloud offerings like AWS. My response to him then was as follows; right now the CEO’s of every major hardware vendor and many software vendors are (virtually) sitting around a table asking each other “how do we fight back against public cloud? The only way is to stay relevant to our customers. The best way to stay relevant is to create solutions that allow internal IT to quickly and cost effectively deploy internal IT infrastructure in a way that closely mirrors public cloud usability.”
The Economist Theories and Cloud or IT use
I’m a big fan of Jevons Paradox, but I only discovered his theories after having written a piece in 2010 on how cloud computing would dramatically increase the creation and use of new IT solutions. I don’t question Jevons theories, because I believe I proved them on my own before discovering his. I also agree with the reference to Ronald Coase. I’m a huge believer in capturing all costs associated with IT projects, and startup time has to be a factor. So from the economist’s perspective I agree with what Bernard has written. What I disagree with is what I would characterize as an over simplification or maybe just an oversight on how IT is used in the big puzzle of creating opportunities with customers.
Why I believe The Private vs. Public Cloud debate is a false one (one of my blogs)
Private Cloud is often compared unfavorably to public cloud, in fact so unfavorably that you can still hear some people say that “there is no such thing as private cloud”. Boy, how would you feel if someone said I think so little of you I don’t even think you exist? I’m guessing it doesn’t get much harsher than that. For and against arguments notwithstanding I’m an advocate for best use of IT, private, public, mainframe, whatever, all that matters is whether it is the right solution for the company and the opportunity at the time.
The assumptions many buyers make about public vs. private cloud revolve around a combination of factors ranging from scalability to ease of use and or ease of adoption. The problem with these assumptions is that they are almost always based on a failed precept that comparing what you have today to public cloud is what should drive your answer. In other words if you compare the legacy environment you have today and how much it costs to run it, replace it, and provision it to a public cloud there’s no doubt that public wins. The problem with that comparison is that it’s a false one. The correct comparison is one that envisions a future state and compares the cost and usability factors of the future state with the long term use of public cloud. It’s also true that real (yes, I said “real”) private cloud is still a baby 12-18 months old, it’s difficult to compare that against the best public cloud infrastructure that is 6-8 years old. In other words, if you’re going to make an assumption about the benefits of public cloud vs. private cloud you need to get as close to apples to apples as you can. Private cloud (per my comment to Bernard in our coffee shop conversation) will be made easier and more cost effective to adopt because there are too many vendors who can’t afford to let anything else happen.
Assuming you’ve made the decision to move to private cloud
How do you compare your use and costs in your private cloud against public cloud offerings? As Bernard pointed out there is the “cost” & “time” associated with acquisition that has to be accommodated, not just the cost of on-going ownership. I agree with Bernard’s assertion in theory, but in practice it doesn’t actually work that way once you’ve made your private cloud investment.
A few very simple scenarios comparing acquisition of compute resources against public cloud
|No Private Cloud (legacy infrastructure)||No comparison||Public cloud wins pretty much every day.|
|Private Cloud Just Acquired (1st use case)||No comparison||Public cloud wins. Just think about it, you buy some converged infrastructure and train/re-org the staff and then provide services or you could go out and buy the services from public cloud|
|Private cloud (2nd use case and beyond)||Very strong comparison||You now have infrastructure and services up and running. Each new requirement for an app can be handled quickly and cost effectively against a pro-rated investment. With each new use case your potential comparison or even cost advantage against public cloud gets stronger|
The above are very simplistic but real world considerations for helping to make a decision on whether to own or lease your compute capacity. The above scenarios are not an end all comparison either, there are dozens of factors (see here) that need to be considered. All I’m really trying to get across is that you can’t compare a one-time use case situation against long term ownership. All too often when we make comparisons we attempt to convince the buyer that the difficulty level will remain the same with each new request and in the case of private cloud that won’t be true, unless you’re planning on building an individual private cloud for each application you own (“that wouldn’t be prudent”).
I’m not a proponent of public, private, or legacy, I’m a proponent for smart purchasing and best fit for the job, the company, the team, and the timing. So think carefully when reading about what the right answer is and instead put it in the context of ‘how will we own this’ and ‘what does done look like?’. Bernard, I think it’s time for another coffee shop talk.