Lately, there has been some talk that suggests that public cloud services eventually cost organizations more than on-premises systems or private clouds.
A recent report in exploreB2B observes how a number of startups shifted to in-house data centers when public cloud services no longer were capable of meeting their growing requirements. One company cited, MemSQL, estimated that Amazon Web Services would have cost about $900,000 over the next three years, versus on-premises servers at $200,000."
NetworkWorld's Brandon Butler also covered this discussion in a recent article, which discussed a report issued by cloud vendors that presents a case study which suggests that "if an organization is spending more than $7,644 in Amazon’s cloud each month, then it can be cheaper to operate a private cloud." However, Butler's article continues, no two cloud formations are alike, and thus this type of threshold is difficult to measure.
Joe Weinman, author of Cloudonomics: The Business Value of Cloud Computing, suggests there are many variables to consider when weighing public cloud versus on-premises systems. The deciding factor, he says, is the consistency of workloads. An organization with highly variable computing needs may find it more advantageous to subscribe to cloud services. If workloads are consistent and unchanging, an in-house system may be the answer.
In a recent chat I had with Weinman, he points out that there's nothing unusual about companies moving off of cloud and back to on-premises systems. This has been going on for some time. There are many variables -- cost is just one -- that determine whether cloud services are the right approach for an organization at a given time.
The cost advantage of cloud services is realized when they are employed to handle spikes in demand, Weinman points out. He provides an analogy of buying a car with $300-a-month payments, versus renting a car at a rate of $50 a day. If there is only a need to use the car once or twice a month, it makes more sense to simply rent it when needed, he points out. "Why would you use the public cloud provider, in this case, the rental car company, versus just doing it yourself? The reason is all comes down to the intermittency of your use, or the peak to average ratio of your demand. So if you need on average one car a day, and your demand is flat, then you may as well as go for $10 per day, which is $300 a month. If hypothetically you only needed the car for one day, then it would cost $50 for the one day, which would work out $50 for the entire month."
There's a tendency to look at unit costs in cloud decisions, which is too narrow a view, Weinman continues. "People fail to realize the value in the cloud pricing model is not the lower unit cost, but the pay-per-use property in the presence of sufficiently valuable demand. So if my demand is flat, I don’t want to pay $1500 a month; I want to pay $300 a month. If its only one day a month or one day a year, then I would rather pay the lower total cost, even with the higher unit cost."
There is another important distinction between cloud and on-premises system that impacts costs, Weinman suggests. Public cloud services have more overhead than in-house IT departments that gets passed on to customers. "An internal IT shop doesn’t need to take out full-page ads in The New York Times. It doesn't need to buy banner ads, it doesn’t need to put together brochures about how wonderful it is. It doesn’t need to pay salespeople. An IT shop doesn't need to worry about all of the sales and administrative costs that a cloud provider needs to build in, and the profit margin that a cloud provider needs to build in, and accounting for uncollectables from customers who don’t pay."
