The Biggest Opportunity is the Network Of Things.
By Ken Birman, N. Rama Rao Professor of Computer Science at Cornell University. October 2014.
Big Data researchers often focus on problems that start after the data has been collected into a cloud computing infrastructure, but I work on networks and infrastructure challenges, where the important questions take a very different form. For me, the really exciting opportunities often center on engaging with the actual sources of all of that data: an entire world of things we might want to sense, track and interact with. What will make this hard is that we’ll need to tackle Big Data problems in which the data is around us, not precollected into big centralized repositories.
For example, I’ve been working on the Smart Power Grid lately. There is a possibility that we computer scientists could write code that will reduce pollution, save electricity and reduce costs by just controlling the power grid in much smarter ways that also make more effective use of renewables. With enough ingenuity the potential is huge: We might even make a serious dent in global warming!
I think topics like this represent must-do opportunities, but they also force us to think about computing in a new way. For example, in the case of the smart grid, we’ll going to have to learn to instrument and control the grid at massive scale. The instrumentation will need to live where the power grid itself lives: out on hillsides and in the woods, exposed to harsh weather and falling trees and curious squirrels. Until we learn to create networks robust enough to deploy into the real world, these kinds of opportunities will remain elusive.
Here are a few challenges that I would love to see people tackle.
- The guy who runs that big network won’t be an expert.
The barrier separating Big Data researchers from the largest opportunities in settings like the power grid, future health-care systems, smart buildings, smart cities, self-driving cars and so forth, frequently isn’t technical at all. The issue is that to operate robustly in demanding settings, we need a new kind of sensor and a new kind of network that can be operated securely and reliably and be installed and serviced by the same people who fix downed telephone or electric wires, or repair traffic lights. If we can make our solutions appealing to those folks, we’ll gain access incredibly high value application areas.
- A lot of that data will be sensitive and it is unethical to collect it at any one place.
I don’t need to tell you how big a problem privacy is becoming. So this is the time for all of us to learn to build Big Data solutions that can do data mining at the edge and do so in privacy-preserving ways. How can a network, and a data mining algorithm, answer a question without compromising the privacy preferences of the end-user? Maybe the answer will be differential privacy, maybe something else, but we urgently need to explore these kinds of questions before the industry just throws something together and deploys intrusive sensing devices that completely lack the needed privacy properties.
- The world is even bigger than you think.
Big Data has already been a huge challenge to us in cloud computing data centers. As the Big Data community engages with the real world, we’ll need to confront the vastly larger reality of data far too large to capture and store in any one place, and real-time requirements far too stringent to permit relaxed, off-line processing. Today’s algorithms will need to give way to a whole new generation of highly decentralized ones that can do a lot of the heavy lifting out where the data is first captured.
But while the future poses challenges, it also is incredibly exciting to think about the potential. I can’t think of a single moment in computer science when we’ve had such rich opportunities to benefit so many people. So if you are a young person thinking about a career in computing, don’t be afraid to jump in. You’ll be challenged, but in good ways, and there is a huge chance to make the world a better place!
– Link to GridCloud