When Multiple Data Centers Make Sense

Although many of the largest tech companies are located in Silicon Valley, often their data is not. With greater frequency, startups and large enterprises alike are seeking to store their data outside of California—looking instead at leasing space in data centers in the Midwest or on the East Coast.

And they aren’t just considering one location either. Savvy technology companies understand that to provide the best application performance and reduce latency, proximity to the end user is critical, which often means leasing space in data centers around the country, if not around the world.

But before we continue, we need to understand at what point a cloud or technology company would consider using a multi-tenant data center and what main factors should be considered when choosing the facility. Enterprise companies may look to outsource data center space when they start to run into challenges such as more-efficient use of space in their own real estate. For example, an expanding employee roster could prompt a company to reassign a modified area in a building that originally housed servers to make room for new employee workspace. The company may also look to a third-party data center operator when a “just in time” need to deploy a new product demands optimal computing space that cannot be built in its own facility. A common occurrence in Silicon Valley is for software companies to build server farms for test-and-development environments close to their engineers in third-party data centers.

On the other hand, a tech startup that is much more focused on application development than on managing its own computing infrastructure will likely find an immediate solution from a cloud provider such as Amazon or Rackspace to host its application. Depending on the critical nature of the application, sharing physical resources in a cloud environment may not be as reliable and secure as building a private cloud in a multi-tenant third-party data center that can deliver a dedicated electrical and cooling infrastructure with high-levels of redundancy, thus offering greater control and lower risks.

After determining that an external data center solution is needed, the questions become where and how does one begin to search. Most companies will look in close proximity to their headquarters. The rationale is that doing so keeps computers close to home-base technical teams in the event of an emergency, such as a server going down. The location would be convenient for local employees to respond to a crisis as well as easily conduct routine maintenance.

Historically, California has been the dominant location for most data centers because of the technology boom in Silicon Valley and the availability of technical resources that could manage the computing platform onsite. But as the market became saturated with rising startups and, by default, data centers, companies started to look elsewhere. California lost some of its appeal for many reasons, including environmental issues (possible earthquakes and drought), higher costs for electricity, real estate, living accommodations and taxes. Improved availability of networks and the advancement of technology to pass data packets enabled further expansion into other geographies without a loss in performance. Places like Oregon, Nevada, Arizona and Washington State jumped at the opportunity to welcome data centers to their municipalities. Located on the West Coast, these areas still offer close proximity to Silicon Valley but add the benefit of lower utility costs, attractive tax incentives from local governments and the opportunity to contribute to improving local economies by building a data center. In effect, California actually helped fuel the expansion of business and jobs outside its boundaries.

A similar trend happened in the northeast. Manhattan is only so large and is densely-packed with financial companies. Available land is scarce and expensive, and what is left is more effectively used for office space since it’s close to the employee population. This trend drove many Manhattan-based companies to move their data centers elsewhere, and New Jersey was an ideal solution. The area had been home to various industrial companies and was rich in fiber-optic capacity, so it became an excellent location to build data centers still within reach of the NY metro area.

End-user location is another factor companies should consider. In addition to the convenience of having a data center close to company headquarters, companies should be cognizant of geographical proximity to customers accessing their applications. Once they determine areas of high customer concentration, it only makes sense for servers and networks to be in the vicinity to ensure quality connectivity and to reduce latency. This is especially true for companies whose core business is transactional, such as the banking industry.

SaaS providers offering social-media sites are great examples of companies that employ a multi-location data center approach for both performance and reliability. These companies know that their customers are always on the go and that they demand the same immediate access to newsfeeds, photos, messages and account data wherever they are and no matter what type of device they’re using. Although customer’s account information may be stored in a physical location close to where he originally signed up, these companies typically have data centers throughout the world so that if a user goes on vacation, he can still view his newsfeed, upload and download photos, and so on—just as easily as if he was at home. In addition to providing high-performance for end users, multiple data centers provide protection from an outage. If one data center goes offline, a customer will be rerouted to another data center seamlessly and with little to no interruption.

So, where are the hot spots for data centers? The United States has a few important Internet exchange points (IXPs) that the bulk of Internet traffic is routed through: Ashburn, Virginia; New York, New York; Chicago, Illinois; Palo Alto and San Jose, California; and Dallas, Texas. Secondary access points include Atlanta, Georgia; Miami, Florida; and Seattle, Washington. The primary focus of the IXPs is to facilitate network interconnection through an exchange access point instead of third-party networks. These exchange points were developed because of population density and provide a common place for Internet service providers (ISPs) to exchange traffic. They are often established in the same city to avoid latency. It’s easy to understand why these areas are popular when you look at their proximity to major population centers and industries.

With IXPs located throughout the country, the market is now wide open, and it’s no longer necessary to only consider the conventional locations. Because Silicon Valley and New York City are two of the most expensive markets, it’s no surprise that the rising costs of electricity and land are forcing enterprises and startups alike to look elsewhere for more-competitive pricing. The cost of electricity in Northern Virginia or Chicago is more attractive than California, as is the cost of land and the cost to build. If there’s no specific reason to be in Silicon Valley, other locations might be more economical.

We’ve come to understand that a company should consider data centers close to its employees and, depending on its business model, close to its customers as well. If multiple geographic locations are needed, it behooves a company to consider a multi-tenant data center provider with locations in multiple markets. Although the factors driving enterprises and startups to seek third-party data center services may differ, the benefits of implementing a strategy across multiple data centers is largely the same. In a world where hundreds of new applications are moving into the digital stratosphere daily, ensuring application availability, resiliency and performance is of the utmost importance.

About the Author

Paul Hopkins is regional vice president of sales and leasing at DuPont Fabros Technology (DFT), where he leads all sales activity in the western region. Paul has been a sales and marketing leader in the data center and cloud services industry for over 20 years. His career began at Netcom, one of the first ISPs and the first Internet company to make an IPO (1994). He then became an early member of the sales team at Exodus Communications, a pioneer in the data center industry whose revenue grew from $50 million to over $800 million in four years. Over the last decade, Paul has built and led several world-class sales teams at leading data center and managed-services providers, including IPSoft, Savvis and Telx. In 2004, Paul earned a dual-MBA from UC Berkeley and Columbia University while working full time and supporting a family. Paul resides in Palo Alto, CA, and is married with two young children.