Xin Sun

Florida International University School of Computing and Information Sciences


Lecture Information:
  • January 23, 2015
  • 2:00 PM
  • ECS: 241

Speaker Bio

Xin Sun is currently an assistant professor in the School of Computing & Information Sciences at Florida International University.
His general research interest lies in networking and systems, and he has worked extensively in the areas of network design and management, cloud computing, and software-defined networking. He was a visiting researcher at IBM Watson research center from June to August in 2014. He received his Ph.D. in computer engineering from Purdue University in 2012.

Abstract

The field of network design is at a crossroad. On the one hand, a decade of sustained research on clean slate network architectures has produced the much-required understanding of why current networks are often brittle and hard to design and operate. Emphasizing (logically) centralized decision making, these new architectures are largely successful at streamlining the building blocks of a network as well as the communication channels between them. On the other hand, these advances in network architecture have for the most part skated the central question of network design; that is, how to orchestrate a distributed collection of devices to meet a diverse set of operational objectives such as performance, security, privacy, and cost. The state-of-the-art of network design remains mostly an ad hoc endeavor, and as a result, a large number of production networks may be suboptimal in terms of the objectives they need to achieve and, worse, design faults are common.

In this talk, I will first present a novel integrated design approach and show how it can minimize the unnecessary configuration complexity in realizing user reachability control, a key network design objective that involves designing three distinct network elements: VLAN, IP address, and packet filter. Capitalizing on newly-developed abstractions, our approach integrates the design of the three elements into a unified framework by systematically modeling how the design of one element may impact the complexity of other elements. As such, our approach goes substantially beyond the current “divide-and-conquer” approach that designs each element in complete isolation. As the second half of the talk, I will sketch out a new research direction that enables enterprises to augment the design of their networks to deliver better services by leveraging the large footprint of cloud providers.