Spiky Demand: However, cloud services may well cost more than some enterprise data centers on a unit cost basis. One might think that this would imply that you should shy away from the cloud, but that's not the case. The key reason has to do with the usage-based pricing paradigm of cloud services. The important insight here is that even if cloud services do cost more when they are used, they cost nothing when they aren't used. This is a very different story than the enterprise data center, where owned resources continue to cost money whether they're used or not. This is the same difference that exists between an owned (depreciated, leased, or financed) car and a rented one. An owned car parked in your garage still carries costs, whether you drive it or not. The key factor in the economics of the cloud is then how spiky demand is. In effect, if you drive your car every day, it's going to be cheaper to own it. If you only need a car once a week, renting is probably better. And, if you only need automobile transport for a few minutes each month, maybe you should just take a cab. If the frequency of use is low enough, it can more than compensate for a potentially more expensive unit cost, and a pure cloud strategy, even if more expensive on a unit cost basis, can still offer a compelling value proposition in terms of total cost.
Any Variability in Demand: Interestingly, while both scenarios above lead to a pure cloud advantage, it is often the case that a hybrid scenario is cost-optimal. Virtually all enterprises have some sort of variability in demand. Retailers have Thanksgiving to Christmas as well as Cyber Monday. Tax preparation firms have a peak in February of early filers and a peak on April 15th of procrastinators. Mortgage lenders are at the mercy of seasonal trends in home buying, shifts in interest rates for re-fis, and macroeconomic factors for home equity loans. We can think of all these firms, however, as having a pretty regular baseline, as well as peaks that rise above this baseline. If cloud services have a higher unit cost, a hybrid strategy is typically optimal. The strategy involves using enterprise data center resources to handle the baseline and cloud resources to handle the spikes, and it's referred to as cloud-bursting. It is often better than using dedicated resources built to peak, which then end up underutilized.
Adjustments to all of these strategies must be made depending on the application and the enterprise architecture. For example, the costs of managing multiple copies of data as well as data transport must be factored in.
Given the importance of this area, an entire track has been dedicated to the topic at Cloud Connect, where we will address both the math behind the economics and the technical architectural implications of that math, as well as empirical benchmarks from a variety of real world examples. To help conduct this analysis, a number of ROI tools and methodologies have now arisen for the cloud, which we will delve into. (Cloud Connect is produced by UBM TechWeb and co-sponsored by InformationWeek.)
Cloud computing is a rapidly changing technology and business model. New dynamic pricing and spot auction markets are arising for capacity, and the future is likely to hold much more in the way of business model and pricing innovation, as well as ecosystem evolution. We will be addressing this topic in depth, and also seeing what the various players in the space are viewing as their advantages and direction.
Cloud Connect promises to be a rich event, and I hope to see you at the ROI and economics track sessions.
Joe Weinman is VP of Strategy and Business Development with AT&T Business Solutions.
Network Computing has published an in-depth report on the state of enterprise storage. Download it here (registration required).