As I've been thinking about cloud computing, public and private, "the long tail theory" keeps coming up. That term describes a business model that sells a few unique items rather than many identical items and has certain parallels to the modern data center.
The idea is based on the Pareto distribution, which when graphed starts high on the left and tapers off into a distinct tail at the right. Long-tail proponents argue that the total sales within the tail in an industry can meet or exceed the sales potential of the newest, most popular items. To me, this sounds precisely like the future of the on-premises data center.
It is fairly common to hear people heralding the death of the data center at the hands of public clouds. Software as a Service (SaaS) will eliminate whole chunks of our data center's workload outright. Platform as a Service (PaaS) will take our custom applications and run them in places where we don't have to worry about common, yet esoteric, things like tuning Java heap sizes. And Infrastructure as a Service (IaaS) is the final nail in the coffin, allowing us to simply move our legacy applications somewhere else.
Applying the long tail theory to data centers
All these services are compelling in certain ways, especially when they address a big organizational problem, such as running out of physical space or a rebellion against terrible traditional enterprise software. They're also compelling when an organization's IT staff isn't particularly skilled in certain areas. Yet the appeal of a private cloud, or no cloud at all, running from a local data center is as strong as ever. This is the long tail on the sales of computing technology, and I wonder if it's large enough to -- pardon the expression -- wag the dog.
Why is there a tail at all? First, there are big issues -- lock-in, pricing and connectivity -- that worry IT staff. How do you get your data out of an application hosted somewhere on the Internet when you find something better? If you've written a custom application against the Google App Engine application programming interfaces, how do you move it when the vendor's pricing, terms or services become unfavorable? With pricing, just as it's cheaper to buy a car than rent it for five years, many organizations have the sorts of workloads that are fairly constant. For them, buying is often a better choice than renting.
The biggest issue with cloud adoption, though, is that you hear of the journey to the cloud most often from large enterprises. And while many small- and medium-sized businesses might use Gmail or Office 365, they see any more cloud as unnecessary overhead. They don't need more IT processes. They don't need self-service, and they're about as consolidated as they'll ever be now that they're virtualized. They may not even need a data center, perhaps just a couple of racks tucked away on their manufacturing floor, the financials of which are just as favorable as cloud computing -- at least for them.
This is the long, happily wagging tail of IT, and it'll be around for a long time.
Dig deeper on Data center budget considerations