Three CCIS Professors Receive Google Faculty Research Awards

CCIS’ Dave Choffnes, Seth Cooper and Rajmohan Rajaraman were granted Google Faculty Research Awards last month for their various research proposals related to technical research in computer science, engineering and related fields.

Applications for the awards are open to full-time professors from universities around the world, according to Google. The awards support applicants for one year, and the median award amount is $50,000-$60,000. On its website for its various research programs, Google predicts that because of an increased number of applications, the acceptance rate for the faculty awards will drop to just 15 percent.

Seth Cooper

“It was very exciting to get selected,” says Seth Cooper, an assistant professor at CCIS. “I know it’s very competitive.”

The project for which Cooper submitted a proposal to Google is based on crowdsourcing image labels. This project, for which Cooper received $61,000, focuses in particular on aerial photographs obtained during flyovers of disaster areas in order to better guide response efforts.

“We’re looking at crowdsourcing the labeling of images, and in particular aerial images of disaster areas, and labeling them based on how much damage there is in the photo,” Cooper explains.

His research at present is focused on improving the engagement and performance of the people who help label the images by examining whether the variety of images provided makes a difference. This is done by automatically clustering the images, then giving different sequences of images to the people and seeing how that affects engagement levels and the accuracy of the labels – “whether you give someone a bunch of pictures of plains over and over and over again, or you give people more of a variety of images like plains, forests, lakes,” Cooper explains.

The next step for Cooper will be to optimize the sequence of images so that the people who sign up to help crowdsource image labels are more engaged and more accurate in their labeling.

“We’ve been using the aerial photography so the idea would be that you could have someone fly over an area with a disaster, get a bunch of images, those could be put into a crowdsourcing system and the labels that are retrieved from the crowdsourcing work could then be used to get an idea of where there’s more or less damage,” he says. “That could potentially be useful in directing response efforts.”

Rajmohan Rajaraman

“Having spent some time there and being somewhat knowledgeable about the space, I know that problems of this kind arise within Google and companies like Google,” CCIS professor Rajmohan Rajaraman says of the project for which he received $58,000 from Google, in addition to $10,000 in cloud credits. Rajaraman’s research is investigating how to improve scheduling policies of large data center networks used by Internet giants, such as Google.

He previously spent a year working at Google, from 2012-13, while on sabbatical from Northeastern. During that time, Rajaraman’s work focused on parallel jobs, or jobs that run simultaneously on multiple machines. A job can be anything from checking a Gmail account to searching for something on Google – it triggers a request that is sent to a server at a data center, and looking at how those requests are distributed and fulfilled is Rajaraman’s focus.

“Some of the work that I contributed led to an improvement in their design and it was eventually implemented and running in all the data centers,” he says. His current work tackles a more complex problem: communication aware scheduling of jobs in data center networks with an emphasis on cloud environments.

Many Internet companies use cloud computing systems to host their servers, which means that jobs are constantly being processed on clouds. Running these complex jobs on clouds means ensuring that there’s high bandwidth availability in case something fails, as well as ensuring that communication between jobs is not hindered. “Many of these jobs could actually be communicating with one another because they might be copies of the same service,” Rajaraman says. “You want to make sure that however they are placed, that communication is done very effectively without any problems or issues. “

Right now, Rajaraman is working on establishing a general framework for tackling this problem. If any companies choose to use this framework, it would need to be tailored to meet the specific needs of that company before going into use.

“A company like Google – they have a massive system in place, and making a change is not that easy,” Rajaraman says. “A lot of things will have to be taken into account before anything like this could actually go into effect.”

Dave Choffnes

“I tend to be drawn to research topics that affect all Internet users, including me,” says CCIS professor David Choffnes about the work for which he received the Google award. Choffnes is working on a project called Meddle, which investigates opportunities to better understand privacy, policies and performance in mobile networks. Google awarded him $58,000 to fund a Meddle application focused on identifying traffic differentiation in mobile networks.

“Differentiation means giving different service to different applications,” Choffnes explains. “It’s a much more neutral way of talking about net neutrality. When we look for differentiation, we don’t necessarily expect your traffic is always going to be worse. In fact, some [Internet service providers] have incentive to make their own streaming services have better performance.”

Net neutrality is the principle that all applications should be given equal access to Internet resources. Choffnes’ work measures when networks provide different service to different applications, with a focus on mobile networks, which are particularly bandwidth-constrained and receive little attention regarding differentiation practices. Videos, which consume significant bandwidth, are especially at risk of being targeted by traffic differentiation policies. In a paper to appear at the Internet Measurement Conference in November, Choffnes’ team found that ISPs slowed down or modified video traffic to reduce their bitrate and consequently their quality – bad news for companies like YouTube and Netflix, and their users.

To measure traffic differentiation, Choffnes and his collaborators at Northeastern and Stony Brook University launched Differentiation Detector, a free Android app, and is working on an iOS app. In addition to the award money, Choffnes received $10,000 in cloud credits from Google, which will allow him to deploy the system for detecting differentiation globally and to make results from the apps publicly available.

“Policymakers like the [Federal Communications Commission], which recently barred many forms of differentiation in mobile networks, could use our system and its data to monitor ISP compliance,” Choffnes says. “I hope this tool not only helps inform users of policies set by their mobile Internet providers, but also becomes a useful source of data for public policy groups and regulators.”