Exploring data center design innovations
Douglas Alger is Cisco’s IT Architect for Physical Infrastructure. He develops architecture roadmaps, solutions and policies for the physical infrastructure of the company’s data centers and other critical facilities around the world. He has participated in more than 80 major data center projects, from all-new construction to substantially retrofitting existing facilities.
In this interview Alger discusses modern data center innovations, the evolution of the data center in the past decade, his latest book – The Art of the Data Center, offers a variety of tips and insight into the future of data center design.
You’ve been to many modern data centers. Which one impressed you the most? Why?
I think each of the facilities The Art of the Data Center are compelling – a site had to be innovative for me to include it in the book. Although the rooms differ from one another, they have one thing in common that’s impressive – they were built in a very purposeful way. The designers and operators had specific ideas about what they want to accomplish with their data centers and they used innovative approaches to do so.
Yahoo designed a computing environment to be cooled without use of mechanical refrigeration, for instance, while eBay designed a data center that shifts from energy efficient to high performance to energy efficient again as customer demand rises and falls. Meanwhile, AISO (Affordable Internet Services Online) built a hosting facility that runs entirely on solar power. Each is impressive, and each is executing on a well-defined vision.
If I have to call out just one facility, it has to be Bahnhof’s data center in Stockholm because it gave me the idea for The Art of the Data Center. I was researching geothermal cooling and underground computer rooms for another book I was writing, Grow a Greener Data Center, and came upon images of the Bahnhof facility. The data center is built in an abandoned, underground nuclear bunker. Its look was inspired by 1970s science fiction movies such as Silent Running and the site has interesting features including artificial waterfalls, a huge aquarium and submarine engines that provide standby power.
After seeing the room, I knew it would be fascinating to speak with the designers and learn what prompted them to build the room how they did.
How has data center design evolved in the past decade?
Data centers today are being asked to do more than before: be more efficient, support more power and cooling density, and foster more productivity – all while providing high availability. The trend began in the early 2000s, when hardware models got smaller and more powerful.
Data center managers could suddenly fit a lot more computing into a given footprint and a server environment’s most critical resource became electrical capacity rather than rack space. Aside from increasing the demand put on data center physical infrastructure, this densification drove up utility bills. Power consumption is now the greatest operational cost of most data centers.
Companies are designing server environments with significantly more electrical capacities than before. Major data centers I worked on 10 years ago involved 50 or 60 watts per square foot, today they have 200 watts and more.
Fortunately, during the past decade there have been efforts across the data center industry to develop more efficient designs. Metrics have been created to evaluate how server environments consume energy, water and other resources, and there are several tools on the market for data center operators to monitor and manage their infrastructure.
Several design and operations best practices have emerged, too, to make data centers greener. Modern designs incorporate higher voltage power distribution, outside air for cooling and isolated hot- and cold-aisle containment and many other strategies to optimize energy efficiency.
What are some of the features a modern data center can’t do without?
A reliable standby infrastructure system to ensure high availability. Monitoring tools to manage system efficiency. Flexible infrastructure to accommodate technology changes in the future. The green best practices I just mentioned, such as isolated airflow distribution. Security measures and change controls to protect hardware and applications from intentional threats and accidental downtime.
Virtualization and unified computing are important, too. People might not think of those as data center design elements, but they have a big influence on a room. You increase CPU utilization and get more computing done through fewer devices, which reduces energy consumption and can allow you to dramatically streamline your infrastructure.
What start-up advice would you give to an organization interested in building a small data center?
Plan ahead. Whatever your initial requirements are for data center space, power and cooling they’ll increase over time. You’ll want to have strategies in place to accommodate future demand.
Understand how you want to use your data center, too. Do you need a highly available, lights-out room that few people enter and where changes rarely occur? Or do you need a dynamic environment where hardware is frequently swapped out and changes happens all the time? Those very different operationally, and you’ll want to design the data center appropriately.
Also, build the room to be as flexible as possible. New technology is going to come along and you want your data center to be adaptable.
What features would you like to see implemented in the data center of the future?
These aren’t necessarily practical, but someday I would like to see:
1. An application that monitors data center infrastructure and shows in real-time how a facility’s resources are being consumed. I want to see power consumption, temperatures and airflow patterns, connections between devices, CPU utilization, occupied rack space and even weight loads on the floor. As new hardware needs to be installed, I want the application to recommend the most efficient place to add future gear – much like a chess program can recommend your best move in a game.
2. The capability to shift processing load among multiple data centers around the world, so energy is consumed at whichever facility has the best conditions at that time. Move demand to a site where it is nighttime and power prices are low or a facility where alternative energy source is temporarily available.
3. Wireless connectivity. Do away with all of the patch cords and structured cabling used in data centers.
4. Computing on demand. Idle servers consume a lot of energy without doing any useful work. Keep those systems completely powered down until they’re needed.
What are some of the interesting facts you discovered while writing The Art of the Data Center?
I’ll touch on three:
Phoenix is a pretty good place for data centers. It’s not the first location that comes to mind for most people because it gets so hot, but two of the sites I profile are there. One of the facilities even employs outside air for cooling, which flies in the face of the assumption that you need a chilly climate to use that technology.
Streamlining data center infrastructure not only cuts costs but perhaps downtime, too. Many companies assume you need more physical infrastructure to achieve greater availability, but more than one data center designer I interviewed said doing away with infrastructure can be helpful because there are then fewer components to fail.
With a bit of planning, you can build a great data center almost anywhere. A concrete silo, former water bottling factory and deconsecrated church are just some of the interesting places where the data centers in my book are housed.