Bringing Internet Speeds into Crucial Hubs of Data Centers and Metro Networks

Achievement date: 
2015
Outcome/accomplishment: 

A novel network switching architecture could enable new speeds and capacity at data centers and metropolitan networks, where much of today’s Internet congestion occurs. The design employs commodity-priced optical space switches (OSS) with software-designed networking (SDN) in a demonstration by the Center for Integrated Access Networks (CIAN), an NSF-funded Engineering Research Center (ERC) with headquarters at the University of Arizona.

Impact/benefits: 

The explosion of video streaming for consumers and virtual computing machines for corporations, among other uses, has placed surging demands on data centers. Facilitating the continued growth of those centers, and the metropolitan-area networks across which they most often communicate, is crucial to continuing those and other uses of tomorrow’s Internet.

Explanation/Background: 

Data centers and metropolitan networks, which cover areas ranging from city blocks to whole cities, are the source of much of the Internet’s congestion. The CIAN switching design has the potential to relieve congestion in data centers and metropolitan networks, where data must be juggled between the optical networks of the broader Internet and the electro-mechanical networks that are commonly employed in commercial and residential networks.

A key to the approach lies in the software, the SDN, which decouples network control from the data stream. That enables dynamic access and much more flexible network control, including the use of optical switches. The SDN was developed at CIAN as Python APIs using TL1 and Binary commands, respectively. The micro-electro-mechanical-system (MEMS) optical switch used in the architecture was developed by CIAN industry affiliate Calient.

 

Network administrators can manage the network from a central console while switch speeds are vastly accelerated with the optical switches. The CIAN design, for example, demonstrated the migration of a virtual machine—software-based servers that are the workhorses of cloud computing—over 50 kilometers with a maximum latency of 361 milliseconds.