Does size matter ?

Does size really matter?

This age-old question has been applied to many topics, but in this case I am talking about Cloud Providers. Everyone knows the usual suspects; Amazon, Microsoft, Google and SalesForce head the lists of Iaas, PaaS and SaaS providers. As in the golden days of the Internet, the idea of First Market Move still seems valid. Once your name is established as the “leader” in any activity, there is a clear advantage in being thought of first when it comes to evaluating solutions.

These days, though, the checkbooks are not opening quite so quickly. Organizations are just getting their feet wet in this whole “Cloud thing.” Is it possible that other new entrants may shock us all with a revolutionary novel idea that eclipses the evolutionary offerings on the market? Or will the next tier of Cloud success stories be "reinventions" of existing old-school players.

It’s clear that the Phone Companies want in on the action. After all, it was the RBOCS and other regional carriers who coined the term “cloud” in reference to carrying data over the Public Switched Telephone Network (PSTN.) They have a pretty sophisticated history of billing customers for usage (if not customer service with a smile.) Phone companies were early adopters of technology, and have a fair amount of experience with their own huge data centers. Why not?

The big Vendors like IBM and HP will certainly want to hang their hats on this new paradigm (especially with server revenue shrinking as a result of virtualization.) In the glory days if selling servers, big data centers would acquire them in lots of threes… one for Testing and Development, one for Quality Assurance and one for Production. The first one might have lower specs, and leave out the HA features (battery-backed write cache and redundant power supplies for example) but usually the QA servers were the same high-end specs as the Production servers. Enter virtualization, and ALL the servers are high end, but customers are no longer buying them in threes. It only makes sense for these companies to “eat their own dog food”, reasoning that of you can’t sell ‘em you may as well rent ‘em out!

IBM, HP and others like Sun and Cisco know how to run a pretty tight data center. They’ve all gone through internal consolidations, and have consultants by the score whom they can leverage. These folks have a lot of experience in High Availability (like SunGard, who is tentatively entering the space as we speak.)

Then, there are the MSPs entering the market, many flying the vCloud flag of virtualization behemoth VMware. What easier way to get a Cloud up and running and give your customers the “arm and fuzzies” than to run your Cloud on the same platform they know and love back home? VMs, after all, are just files, and don’t know or care where they run. More than likely, Cloud will follow the same trajectory as virtualization did over the past decade; OK for Test and Dev / Q&A but not for production until it is proven.

Perhaps less-critical services will move off-premises and these players will pick up the slack for seasonal demand surges where security is not a big issue. Once they have proven themselves, more and more Core applications will migrate outside the Corporate Data Center.

Two related issues are scale and cost. How can a minor player provide that holy grail of “illusion of infinite elasticity?” How can smaller providers compete with Internet Giants who own enormous data centers, buy their electricity on Futures markets, and hire the smartest engineers in the world to keep their Open Source-based infrastructures humming? That will indeed be a challenge.

My friend and mentor Bernard Golden, who covers CC for the prestigious periodical CIO Magazine, did a very enlightening comparison between AWS and a typical VMware VCloud Express provider:
http://www.cio.com/article/502886/How_to_Compare_VMware_and_Amazon_Cloud...
The numbers are not SO far apart, that one cannot imagine as competition heats up, and more players come up to scale, both approaches will be scrutinized before P.O.s get signed. Certainly the larger players have the upper hand when it comes to scale, but a little capacity planning can go a long way. It is a luxury to spin up and down a thousand servers at the flash of a credit card, but the ability to work with the files and APIs you already know is quite tempting. Unless you are the type of organization whose server needs fluctuate so wildly that you cannot predict a demand change in the order of magnitude accommodated only by the giants, a smaller player may be the right choice.

Perhaps the old phrase “one-size-fits-all” doesn’t apply here.