Amazon Adding Cloud Capacity in Northern Virginia

amazon-racks

A rare view of the server rooms of an Amazon Web Services data center, from a presentation at an Amazon Technology Open House last year."

Amazon's move to lease two new data centers could shore up the reliability of its cloud computing infrastructure in northern Virginia, which hosts key facilities in the outage-plagued US-East region. Amazon is reported to be the anchor tenant in a new data center project in Ashburn, Virginia that was disclosed this week.

On Monday, Corporate Office Properties Trust (COPT) confirmed that it will build two large data centers in Ashburn's "Data Center Alley" that are fully leased to "a subsidiary of an investment-grade Fortune 500 company." The Washington Business Journal has identified the tenant as Amazon.

COPT will soon begin construction on a 200,000 square foot data center, which could come online as soon as the fourth quarter of 2013. The developer expects to begin construction on a second 115,000 square foot server farm building by mid-2014.

The new facilities are the latest in a series of data centers Amazon has deployed in Northern Virginia, one of the Internet's key intersections. The US-East region has been the focus of multiple outages for sites hosted on Amazon Web Services, including two outages in June, and downtime in October and on Christmas Eve.

Oldest Facilities in the EC2 Cloud

The US-East region was created in 2006, and includes data centers that were opened as early as 2004. In addition to being the oldest part of Amazon Web Services' infrastructure, it is also the most crowded, according to an independent analysis that suggested up to 70 percent of the IP addresses at AWS were housed at US-East.

Although older infrastructure isn't necessarily less reliable than newer facilities, data center design has evolved over the years to allow new facilities to support higher power loads and denser server configurations. This is a consideration for Amazon and other cloud computing providers, who seek to make the most out of their server and data center capacity, making their facilities more highly utilized than traditional data centers.

Can additional data center space help improve uptime for US-East? Amazon doesn't release a lot of detail about its data center infrastructure, which is split into nine regions, each with multiple availability zones, allowing customers to create redundancy for key applications. The US East-1 region consists of more than 10 data centers structured into five Availability Zones, Amazon Web Services said in an incident report following its June 29 outage. "These Availability Zones are in distinct physical locations and are engineered to isolate failure from each other," the company said.

So why do problems keep occurring in US-East? Why is this region so crowded, and why do sites like Reddit, Heroku and Pinterest continue to host infrastructure in US-East? Barb Darrow outlined some of the reasons for this in a recent post at GigaOm, including the availability of new features (which roll out first in US-East) and the challenges for long-term customers in moving apps and large volumes of data between regions.