Vincent Liu is Networking the Cloud

In today’s computing environment, the vast majority of tasks are inexorably linked with cloud computing. “Whether people realize it or not, much of today’s data and computing, including online services such as website hosting, maps and navigation, social media platforms, streaming media providers like Netflix, Spotify and YouTube, and file storage platforms like Dropbox and iCloud, are all actually hosted in the cloud,” says Vincent Liu, assistant professor in the Department of Computer and Information Science. As a result, efficient cloud networks are becoming ever more critical to ensuring the smooth operation of the vast majority of what is happening on your desktop, laptop or mobile device.

Liu is a researcher with broad interests in network computing. His work bridges all layers of the networking stack, from hardware concerns to application and user demands. After receiving his doctoral degree in Computer Science from the University of Washington, where his research was in improving the functionality of data center networks, Liu spent the fall of 2016 at Facebook, where he was able to work closely alongside the team that designs their network systems.

“Facebook runs one of the largest computer networks in the world, and I got to sit with the team that builds the software systems that run that network,” states Liu. “Being there taught me a lot about the challenges that accompany building networks of that scale, and the priorities of the people that work on those networks. It also allowed me to get my hands dirty working with real systems. I learned a lot from the experience.”

Though it may be tempting to picture “the cloud” as a floating, intangible thing, in actuality it is more helpful to think of the cloud as a type of distributed computing — all of the cloud is still physically stored on networked computers. “Take Gmail for instance,” Liu continues. “Your emails are spread across several geographically distributed warehouses that each contain tens to hundreds of thousands of computers — the cloud. Leveraging these ‘pools’ of computers is much more cost effective than storing it on your own hard drive, and it also allows you to access it anywhere, anytime, from any device, and with little fear of ever losing your data.”

All of this dependence on the cloud creates an increasing need not only for research in the areas of managing big data, but also in network system efficiency, reliability and security. This is where Liu has found his research interests leading him. “What I work on is improving the science behind the network that connects those tens to hundreds of thousands of computers,” he states. “The network is an essential piece of the puzzle, as it’s what allows all of the computers to work together and then deliver data to and from users.”

In addition to his current work in networks, Liu has followed research questions down many related avenues of computer science. “To borrow a phrase from one of my Ph.D. advisors, Tom Anderson,” Liu states, “‘I see myself as a generalist. I am attracted to the biggest problem I can find, regardless of area.’ To this end, my projects have touched on fault-tolerance, security, data centers, wireless networks, clean-slate internet architecture, routing/addressing, and the economic aspects of the internet. My undergraduate research was in the area of compilers and parallel systems.”

Security concerns are most certainly at the forefront of users’ minds when they think about so many of the applications they now use being cloud-based. But Liu makes a distinction between concerns about malicious attacks from hackers versus user privacy when it comes to using the cloud.

“I actually think that the cloud is the answer to those security concerns,” says Liu. “Now, it’s certainly the case that we all have more personal and valuable information in digital form than we used to. This makes breaches more serious, but compared to keeping the same information on your personal computer or trusting each individual company, the cloud is generally much more secure — cloud applications and infrastructure are typically both better defended and more up-to-date. The thing that cloud computing is much worse for is privacy. Until earlier this year, using Gmail meant that Google could ‘read’ your email and serve you ads based on the contents. Privacy is a particularly complicated problem that will probably involve some combination of technology and policy, but I think the first step is educating people about the privacy of their data.”

Liu became interested in computers at a very young age. “I love figuring out how things work, and I also love building things,” he states. “Computers offered me an accessible way to explore both of those passions.” He earned bachelor’s degrees as a Turing Scholar in Computer Science and a Dean’s Scholar in Computer Science at the University of Texas at Austin. He then went on to the University of Washington, where he obtained master’s and doctoral degrees in Computer Science & Engineering.

Liu was motivated to come to Penn because of its collaborative environment that supports cutting-edge work in his areas of interest. “Penn seemed like a place I would really enjoy working,” he recalls. “In terms of research, there are quite a few people here that are working in areas related to computer networks, so I thought my skillset would be a good complement to the rest of the faculty. Moreover, it seemed like the department had a culture of mentoring and collaboration. In terms of teaching, the students here are clearly very strong, and I knew it would be a pleasure to teach them.”

His interests in networking with respect to cloud computing will also help inform what Liu teaches his students. “I’m definitely biased, but it’s hard for me to think of anything I do on my laptop or smart phone that doesn’t eventually depend on the internet or the cloud in some way. Because of this, I’d liken learning about cloud computing to learning about operating systems or computer architecture or compilers — students are learning how computer systems work, the principles behind their design, and how to use them more effectively.”

Looking forward, Liu hopes to focus his research on challenges that stem from the rapid growth of cloud networks. For example, as more things increasingly depend on the cloud, the effect of network outages increases as well. Another issue is that the scale of these networks are challenging a lot of the assumptions that network architects have made over the years. “In response to these challenges, I’m interested in investigating how to limit the effect of these outages,” he states, “and I’m looking at various ways in which we can design networks that are more suited to cloud environments.”

Liu will have ample opportunity in the coming years to explore his interests and to identify and answer new questions. The offloading of data and computing to huge warehouses of computers has enabled things that would not have been possible otherwise, and the size and scale of these deployments make computing cheaper, more flexible, and more reliable than ever before. “Unfortunately,” he states, “there are components to these systems that fundamentally do not grow at the rate we would like, such as the power available at any given location and the latency from these relatively centralized locations to customers. We need to figure out what cloud computing and cloud infrastructure looks like as they continue to grow in size and importance.”