CS Colloquium - Semantics-based Network Management: in Search of a Matching Structure
Network management used to be a black art, requiring the admin to manipulate bags of distributed low-level protocols (the structural backbone) to achieve an (semantic) intention, often by trial-and-error. With the rise of software-defined networking (SDN), centralized and high-level programming abstractions have emerged, forming a new network structure that is easier to operate and amenable to formal verification that provides stronger semantic guarantees. Is the combination of SDN structures and formal verification the long-awaited solution for network management?
In this talk, I will examine the mismatch between state-of-the-art network management in the SDN era and the unique demands of networking as an ever-evolving, shared, and massive-scale infrastructure. Informed with this analysis, I will present my exploration with relational structures from databases to enable semantics-based network management, addressing pain points such as coordinating multiple stakeholders and reasoning with partial information. I will also discuss how relational structures might allow us not only to verify network semantics, but also to explain why and how.
Anduo Wang is an assistant professor at Temple University. Before that, she was a postdoc at UIUC with Matthew Caesar and Brighten Godfrey. She received her PhD from UPenn in 2013. Her research improves network state management with formal methods, databases, knowledge representation and reasoning, logic programming, and more, funded by three (single-PI) NSF grants (CRII in 2017, CNS core small in 2019, CAREER in 2022). She served in various networking venues (e.g., ACM SOSR general chair, 2020) and the ACM SIGCOMM Executive committee (SIGCOMM Information Services Director, 2017-2021).