Interview with Jim Smith, CTO of Digital Realty (http://www.digitalrealty.com/):
DCP: Jim, you’ve spoken about modular datacenters and other next-generation datacenter designs at recent conferences, and that’s a topic that a lot of our readers ask about. Is ‘modular’ just a buzzword, or is it something that is going to have a real impact?
Smith: I definitely think it’s real because it has significant measurable benefits in terms of construction costs, reliability, energy efficiency, operating costs and more. Modular isn’t a passing fad. In fact, I feel strongly that it will become a cornerstone principle for how datacenters will be designed and built going forward.
With that said, I should clarify something. Modular datacenters, to me, are not the same as “containerized datacenters” or “pre-fabricated datacenters”. Some people use those two terms interchangeably with modular, but to me they are very different. Containerized or pre-fabricated datacenters are very cool pieces of modern datacenter design, and they have very interesting applications. But those solutions are not what I mean when I talk about modular. When I say “modular” I mean brick-and-mortar datacenter buildings that have all the benefits of a solid, permanent structure….but built with pre-constructed components that translate into higher reliability, lower costs and dramatically faster completion timelines.
There is a quiet revolution happening in the way datacenter buildings are being built, and modularity is at the heart of that. The completed buildings may look similar on the outside and the raised floor space may look just like finished datacenter space always looks, but the way you got there is radically different and a huge step forward.
DCP: Can you talk more about pre-construction? What kinds of components are pre-built?
Smith: Pre-building is a big part of taking a modular approach to building datacenters. And you’re right, we’re not pre-building entire datacenters, driving them down the highway on a wide-load trailer and plucking them down on a cement pad. We call our new datacenter design and construction process POD Architecture 2.0, and it is the logical evolution of the modular construction approach we pioneered with our original POD Architecture.
We have selected key components of the datacenter to pre-build offsite in a controlled, clean, highly-efficient factory setting and then deliver them exactly when they are needed in the construction timeline. Electrical rooms and mechanical/cooling systems are the most important components we are pre-building and inventorying, and we chose those for a number of reasons.
The primary reason why pre-building is better is how much it speeds up the process. Any of your readers who has been involved in a datacenter build knows that certain steps in the workflow are liable to be bottlenecks that slow down the entire project. On-site construction of the electrical rooms and cooling systems are two of those potential bottlenecks because of their complexity and how they sit in the critical path of other steps in the construction process. By pre-building these components off-site and delivering them at pre-determined points in the construction process, you keep the timeline on track and avoid costly, frustrating delays. Plus there is a higher level of quality control in a manufacturing environment versus a construction site, which results in a higher quality, more reliable facility.
DCP: When you hear about datacenter builds going millions over budget and weeks or months over schedule, are these common culprits?
Smith: Construction projects in general are susceptible to cost overruns and delays, as everyone knows. That’s the nature of construction. But datacenter projects are particularly vulnerable because of the highly technical nature of the project and the huge amount of capital involved.
We’ve all heard horror stories about projects that have gone awry. Sometimes it’s just poor planning. Sometimes it’s that the company keeps changing its mind about what it wants. Sometimes it’s bottlenecks caused by those complex steps in the build. Those problems have existed for as long as folks have been building datacenters. Through POD Architecture 2.0, we are addressing those long-standing datacenter challenges in ways that reduce costs, reduce risks, accelerate construction timelines, improve quality and achieve higher energy efficiency.
DCP: In one of your presentations you talked about how much time a modular approach shaves off the construction timeline for a datacenter. The numbers are almost hard to believe.
Smith: The difference is dramatic, and it is pretty stunning when you first hear it for the first time from industry folks like me who are sharing case studies. But when you walk through it, it makes perfect sense. Using POD Architecture 2.0, we can deliver a datacenter in as little as 12 weeks. The industry average using traditional design and construction methods is in the 9-18 month range. That’s not only a fraction of the time, but millions of dollars or tens of millions of dollars in cost savings.
That stuns a lot of people when they first hear it, but everyone who has been involved in a datacenter build knows the most painful steps and the causes of the biggest delays. They know how many weeks can be lost on bottlenecks. They know the problems and common delays with traditional construction processes. And this next-generation design and construction methodology simply goes after each of those problems and re-thinks them in a way to make the entire process simpler, faster and less expensive. When you start removing those most common delays from the equation, the math takes care of itself and these dramatically shorter project timelines make perfect sense.
It appeals to me because it is highly conservative from an engineering point of view. This isn’t radical design. We are simply catching up to what other industries have known for decades: that quality goes up, costs go down and product time drops significantly when you utilize modular best practices.