Posts

Akshay Narayan is skeptical of absolutes. A postdoctoral researcher at the University of California, Berkeley and the International Computer Science Institute, his work focuses on ways to make specialization for network environments accessible, and this fall, he joins Brown CS as assistant professor. He’s one of the four latest hires in the multi-year CS With Impact campaign, our largest expansion to date. 

“There’s a divergence,” he says, “between the abstractions we use as undergrads writing apps that use the Internet and the hyper-optimized level of organization in data centers. The two viewpoints are somewhat in conflict, and a lot of my research has been about asking the question: there are these two competing perspectives, can we reconcile them? Is there a different way?”

Although generally interested in gadgets and computers from his early youth, Akshay explains that he isn’t one of those computer scientists whose coding experience started at age three. As a Berkeley undergraduate, he initially expected to focus more on the engineering side of a joint computer science/electrical engineering major.

“But I took an intro computer science course,” he says, “and I was hooked. At the time, cloud computing was taking off, and the idea of data centers running your code at massive scale was very interesting to me. In my third CS class, my professor gave us Amazon EC2 credits to have a homework project run in the cloud, and I remember how cool it was to write commands and have a huge computer somewhere else do what you want it to do.”

After working with Professor Sylvia Ratnasamy on resource disaggregation in data centers, Akshay remembers asking, is there more of this?  What he calls a “gradual narrowing” of his research interests was taking place, and it continued over a Master’s degree and a PhD at the Massachusetts Institute of Technology: “The simple idea that memory is going to be moved far away from compute, and our obligations have to adapt, that definitely informed my way forward.”    

Currently, Akshay tells us, his work pursues two avenues along the broader theme of reconciliation. 

“First,” he says, “is rethinking the way we build apps. In the traditional model, we assumed that a best-effort network will move our data. How your app’s experience is conveyed to users is very important, but now, with the cloud, there’s all this extra processing that we really need to worry about. Everything is very complicated, with massive teams of engineers on rotation when something breaks in Southeast Asia and someone gets a call in the middle of the night. Is there a different way that we can build apps without so much vertical integration? What should the app think the model of the Internet is? How do you make sense of all these different cloud services and providers?” 

The other avenue is focused on the way that people worldwide share the Internet’s resources: “We have this intuitive understanding that we’re sharing the Internet with other people, but how much resources do you get, does a researcher at Brown get, does someone else get? There’s no principled way that Internet resources are allocated to various users. We have these protocols to avoid overuse, bandwidth allocations, but what we’ve seen more recently is a move away from this. Is this traditional mechanism still the mechanism by which we share resources today or is there another? Can we design one with more predictable rules that give more equity to individual players?”

Akshay is looking forward to answering these questions with what he describes as Brown’s “strong and passionate” undergraduate population.

“Students here are very different,” he says. “There’s a passion for research at an undergraduate level to a much greater extent than I’ve observed elsewhere. Brown has people who really want to get their hands dirty with research and move the field forward. Working with students like that is always an interesting prospect.”

On the subject of passion, the first thing that motivates his work, Akshay says, is simply enjoyment: is this problem fun to work on, and am I going to be working on it with people who are fun to work with? But as he describes his second motivation, he returns to the idea of balance and reconciliation: “How is the thing that I think is fun going to be useful to people? Is it going to make the Internet more distributed or more centralized? Does the average person have as much access as anyone else, or do some have more than others? A distributed Internet isn’t bad, and a centralized Internet isn’t bad, but these and other axes should be balanced. The solution is some middle ground.”

When looking for fun in his personal life, Akshay describes himself as a casual sports fan (Formula 1 racing, soccer), a baker of sourdough bread, and an outdoor enthusiast. Having recently become a biker, he’s excited to try sailing after moving to Rhode Island. He says that they’re distractions, not always successful, from the sometimes over-full life of an academic. “In theory,” he jokes, “physical activity is very relaxing.”

Coffee is a major interest, and Akshay maintains a list of some of the best places to drink it worldwide. The ritual of its preparation is appealing, he says, as well as the enjoyment of taking a break, but there’s an experimental aspect as well, an intellectual challenge: “It’s an optimization question. With very similar inputs, before you even buy a cup of coffee, there have already been so many different steps of processing, so much done that alters the output, and each affects how your coffee tastes. That interests me.” 

We end our conversation on a humorous note that harks back to the theme of reconciliation when I mention seeing a quip on social media that Akshay had reposted: “What if I like doing stuff that is tedious and error prone?”

“I think a lot of systems research,” he says, “is the idea of traditional abstraction versus modern hyper-optimization. A lot of work has been about building abstractions to hide tedious, error-prone tasks or to check them, to make sure it’s not possible for you to shoot yourself in the foot in the first place. So there’s a classic joke that systems people will do tedious things so others don’t have to, but a lot of what motivates me is that we need someone who can look at two opposing interests and find a different way. For me, that isn’t tedious at all.”

For more information, click the link that follows to contact Brown CS Communications Manager Jesse C. Polhemus.