Wisconsin Alumni Research Foundation

Aditya Akella

Man behind machine
Aditya Akella has found a niche for his Stratos project in the gap that exists between cloud computing architecture

Stratos project a high calling for computer science leader

Aditya Akella considers himself a high tech plumber. And it likely won’t be long until tech-savvy consumers everywhere are grateful for his efforts to keep data flowing smoothly through the virtual pipes and other fixtures that make cloud computing possible.

Akella, an associate professor of computer sciences at the University of Wisconsin-Madison, understands he has chosen an unexpected route to technological achievement. While accolades rain down on software developers responsible for the latest smartphone apps and social networking tools, the technological underpinnings and algorithms that speed, protect and direct the flow of data receive far less attention.

An expert on these linkages, Akella and his team find their calling in advancing the technology. To fully achieve the promise of cloud computing—accessing computing resources including storage and applications from the Internet rather than local servers—improved “middlebox” connections will be necessary.

“Large companies and entrepreneurs alike have focused long and hard on advancing computation and storage technology,” Akella says. “By comparison, network infrastructure is still in the Stone Age. The traditional view has been that it’s akin to plumbing. The reality is, without faster and better networking tools, the advantages of cloud computing ultimately will be limited.”

Among the keys to network advances are the “middleboxes,” devices that transform, filter, route or secure data. Examples of middleboxes include firewalls, which filter out dangerous or unwanted information; wide area network optimizers that coordinate to cache or compress data moving across the Internet; and load balancers that help route the movement of data to multiple access points.

Akella’s team is developing a softwarebased framework called Stratos that makes these and other network middlebox services as flexible and efficient as the applications they enhance. Stratos automatically determines how best to deploy middleboxes in the network architecture, intelligently balances and directs traffic to avoid network congestion and scales the middleboxes to optimize performance and cost.

The software allows users to run middleboxes developed by other vendors and seamlessly strings multiple technologies together.

“Without Stratos, users have to determine how many middleboxes they need and manually configure traffic, a setup that is never optimal because the volume of traffic can change,” Akella says. “With our system, when traffic picks up, it automatically identifies the most beneficial configurations and employs the resources necessary to handle the load.”

Thus, large cloud service providers could employ Stratos to enhance services to customers while increasing their own efficiency.

What makes Stratos work? More than 10,000 lines of code created by Akella and student researchers Aaron Gember, Saul St. John, Anand Krishnamurthy, Robert Grandl and Xiaoyang Gao.

Akella credits his experience with WARF’s Accelerator Program for giving him the direction needed to push forward and complete a working prototype of his system.

“The Catalysts told me to go ahead and bring on some students to get it written, rather than continue trying to do it all myself,” Akella says. “This was really critical because without the added support, I would still be balancing this project against other responsibilities and progress would have been much more limited.”

Instead, Akella’s working prototype already has caught the attention of several major potential industry partners. He has provided live demonstrations to Cisco Systems and Dell and estimates he needs just a few more months to fully scale up and debug the team’s work.

“Industry is just now coming to grips with the expanding needs in this area,” Akella says. “We see the potential to create value for society in a number of ways. Our research group has a tradition of releasing free, limited versions of new software to other academic groups for research purposes. We also intend to create a version with more user-friendly interfaces for commercial purposes and license to industry or form our own startup company.”