Carnegie Mellon University
Postdoctoral Researcher
Guyue (Grace) Liu is a postdoctoral researcher at Carnegie Mellon University, working with Prof. Vyas Sekar. She received her Ph.D. in Computer Science at the George Washington University in 2019, advised by Prof. Timothy Wood. Her research interests are in the areas of networking, security, and systems. She has published research papers in top conferences in the field including SIGCOMM. She has interned and collaborated with leading research institutes, such as Microsoft Research and Hewlett Packard Labs. She received a bachelor’s degree from Beijing University of Posts and Telecommunications in 2012. She has won the HP Helion OpenStack Scholarship, the First Place in GENI Competition, and an RTAS Best Student Paper Award. She was selected as one of the 10 N2Women rising stars in networking and communications in 2019.
Research Abstract:
Novel Abstractions for Software-Based Networks
Network Functions (NFs) (e.g., firewalls, proxies, and caches) are crucial to modern networks to meet security, performance, and policy compliance goals. Specialized hardware devices used to perform these tasks but are no longer a good choice for cloud providers and large carriers due to the high cost and low flexibility. Thus, network operators turn their gaze to software and ask is it possible to run network functions on commodity servers? An answer requires processing packets at 10 Gbps or beyond to avoid slowing down applications; conquering software failures without disrupting network operations; managing large-scale hosts efficiently to minimize extra communication overheads–yet traditional software and virtualization techniques can’t fulfill these requirements.
I work at the intersection of systems and networking, and my dissertation answered this question by building Network Function Virtualization (NFV) platforms, demonstrating not only how to design and optimize the underlying system to overcome performance problems but also how to provide flexible abstractions for applications, hiding system-level communication and processing concerns. More specifically, I have designed (i) a high-performance service chaining platform that provides a unified abstraction, enabling packets to be moved efficiently at each level: from the NIC to the NF, between different NFs, and across different hosts; (ii) A novel service chaining abstraction that breaks the monolithic model into modular micro-services. It includes a group of customizable micro-stacks and convenient higher-level micro-events interfaces, which effectively eliminates redundant processing and simplifies programming; (iii) A distributed real-time performance monitoring platform that allows users to unobtrusively identify and extract necessary data in the middle of the network, without instrumenting the application or understanding the complicated underlying network topology.