Interesting question about building a new data center (or not building one)…

A great conversation I was a part of that still has me thinking. Given the state of IT today, in the send that cloud is an option at what point does a company no longer build their own data center.

The first point would be a startup probably won’t build a data center. Renting a cloud solution and building a VPC (virtual private cloud) is in the long run a much better solution. AWS and Azure are the ultimate incubators for new companies. Free shared service IT provided. Well not free but not at the cost of bringing in a 24/7 IT staff.

Beyond that though when would a startup move to colocation or even building their own data center. Netflix never has, they continue to grow and continue to use AWS. So given that I suspect most startups today may never move to either a co-located data center or build their own data center.

This brings me back to the question we discussed. At what point would a company need to make a decision?

First Criteria: If the cost of the overall cloud solution was more than it would cost for them to build, power, cool and staff their own data center. Here is where the concept of a cloud container Data center – flexible moveable and ultimately expendable may play.

Second Criteria: The usual suspects – unique hardware requirements. Your solution requires a physical connection to a specific piece of hardware. The thing about the IoT is that the direct hardware connection is becoming the dinosaur of computing. There are fewer and fewer of these creatures left on the earth.

Third Criteria: Security. Although I would argue that security has to eventually become literally physical security required. Cloud security which is the most bemoaned problem of cloud computing isn’t as bad as the experts say. The reality of security is anytime you have the potential for human error you have a security risk. One cloud professional managing 1000 servers in the end as long as all things are equal, may be less risk than 5 on-site admins.

I am sure there are more criteria to consider and I suspect I will get emails with more over the next few days but this is my starting point. What is in the end (sorry Mr. Gladwell) the tipping point for a startup to move out of the cloud and into a co-located data center or their own data center?

The other side of this is of course the reality for a company that already exists. A company that has data centers today but wants to build a new facility. Do they cloud the data center, co-locate the data center or build their own. The tipping point for an existing organization is much different than for a new organization. Simply because cloud offers the new or start-up company the opportunity to get up to speed with infrastructure and software services quickly. What once required the purchase of networking, servers and software can now be done in the cloud quickly (still of course buying or renting the software). For an existing company they run into the broader issue of cost.

It was a great conversation and one that I am sure will continue.


Scott Andersen

IASA Fellow