Da Yu

115 Waterman Street (CIT Office: 319)
Department of Computer Science, Brown University
Providence, RI 02906
Email:

dyu [at] cs.brown.edu

I'm a 2nd year PhD student, advised by Prof. Rodrigo Fonseca. My research focuses on Computer Networks, Distributed Systems. Now, I am mainly working on Software-Defined Networking (SDN). I am interested in modeling and understanding the interactions between high-level controller applications and the network hardware, in both directions. More specifically, I am interested in ways to verify and assert that the network is behaving in the ways specified by the controller software, and also in synthesizing controller software to perform in a way similar to an existing (traditional) network.

More Details About Me: CV.

140x140
  • Apr - 2015: Our paper: "Simon" and "Exodus" are both accepted by SOSR 2015!
  • Apr - 2015: I will have an Intenship with Alvin Auyoung at HP Labs in this summer.
  • Apr - 2015: I will give a talk in ONUG 2015!
  • Apr - 2015: I will attend NSDI 2015!

Exodus: Toward Automatic Migration of Enterprise Network Configurations to SDNs

Exodus is a system that consumes a collection of router configurations (e.g., in Cisco IOS), compiles these into a common, intermediate semantic form, and then produces corresponding SDN controller software in a high-level language. Exodus generates networks that are functionally similar to the original networks, with the advantage of having centralized programs that are verifiable and evolvable. Exodus supports a wide array of IOS features, including non-trivial kinds of packet-filtering, reflexive access-lists, NAT, VLANs, static and dynamic routing.

Simon: Scriptable Interactive Monitoring for SDNs

Although Software-Defined Networking can simplify network management, it also poses new testing and debugging challenges for operators. SDN developers need to be able to inspect rules in switches, correlate rules across the network, and verify that the rules installed in switches correctly produce the expected high-level network behavior. To address these challenges we are developing SIMON, a novel scriptable interactive monitor and debugger for SDNs. SIMON allows programmers to detect bugs by describing what the network should do (an ideal model), without necessarily having deep knowledge of the controller application. Using SIMON, programmers iteratively refine their understanding of the system at a debugging prompt, similar to traditional debugging tools. SIMON enables, for example, the automation of repetitive tasks and the monitoring of invariants. SIMON is ongoing research and our initial results on test networks are extremely promising.