Title: SIGCOMM from the Past: Stories & Experiences
Host: Behnaz Arzani(Microsoft Research)
Scribe: Ziyi Wang (Xiamen University), Kexin Yu (Xiamen University)
Introduction
Linxin Gao discusses how to achieve stable inter-domain routing in a decentralized environment like the Internet, where no global coordination is required. Drawing from her curiosity about how Internet routing works, she delves into the limitations of the Border Gateway Protocol (BGP), including the persistent routing oscillations it can cause.
Q1: Since the Internet routing system is so complex that it is difficult for us to fully understand its internal mechanism, how can we perform effective verification?
Lixin Gao: This is indeed a very challenging problem. We are investigating how to make use of the existing hierarchical information to speed up the verification process of BGP policies. Not simply exploring the root node, but better understanding how the root node is formed can help speed up the verification.
Q2: How do you feel about the results of your research appearing in a dialogue system like ChatGPT?
Lixin Gao: I never thought my research would have such an impact. We should not think too much about the impact of the research results but rather focus on answering the questions themselves. My attitude is to focus on exploring the question rather than focusing too much on the impact of the research.
Kenneth Calvert outlined one of his early research efforts on protocol conversion for computer networks. The problem of interconnecting the many different network protocols that existed at the time was described, and the researcher proposed a formal model based on finite state machines to derive a protocol converter that met the requirements. This work took place at a time when the TCP/IP protocol was not yet fully dominant in the marketplace and reflected the academic community’s search for solutions to the technical challenges of network interconnection.
Q1: What is your understanding of the approach to converting HTTP to ICN (Information-Centric Networking)?
Kenneth Calvert: I indicate that I have seen an outline of an approach to convert HTTP to ICN, but it was not a formal approach. I recall it being from a company, but I am not certain of the exact source. I suggest that using solvers and formal specification techniques may be worth looking into for this type of protocol conversion problem, given the complexity involved.
Q2: What is the history behind the best paper award at the SIGCOMM conference?
Kenneth Calvert: According to the discussion, the SIGCOMM conference used to only have a best student paper award, and took pride in that tradition. However, about 15 years ago, they decided to start giving the best paper award as well, but would only award the student paper prize if the best paper was not eligible for it.
Minlan Yu presented her early research on Software Defined Networking (SDN) during its nascent stages. She explored how to achieve fine-grained network control while dealing with limited controller resources and coordination with switches. She proposed a hybrid approach combining proactive rule installation with on-demand rule installation to address the issues of limited switch space and controller processing delays. Through experiments conducted on the Nox controller, she demonstrated the scalability of this approach in handling large volumes of traffic.
Q: Minlan, you were working on software-defined networking (SDN) at its inception, much like you were involved in the early stages of the commercialization of the Internet. At that time, there were many areas waiting to be discovered; it was a completely open space. Did you feel that way while doing the research, and how does it feel now?
Kenneth Calvert: TCP/IP was beginning to show that it would dominate the world, but there was no web, no World Wide Web at that time—only FTP and email. We didn’t realize how crucial architecture, end-to-end principles, and operations at the edge would become. What strikes me now is the importance of the ecosystem; the flow of funding significantly influences what we can and cannot do. I think there is still a lot of work to be done in this area.
Lixin Gao: I would say to younger people that you should really recognize the resources you have now. You can achieve much more than I could at the time.
Minlan Yu: At the time, I was still a graduate student and did not realize how significant this trend was. It was only after extensive work that I started to understand its importance. From the perspective of reviewing the SDN trend, we did grasp the problem—that management was difficult and needed solutions. However, after the community tried many different approaches, we found that the most influential product was actually at the edge, not inside the switches. Many of our small design decisions were incorrect. This indicates that understanding the problem is even more crucial than the subsequent work, as it pertains to the larger direction.
Personal Thoughts
As a student, I am deeply moved by the theme “SIGCOMM from the Past: Stories & Experiences.” Reflecting on and discussing the experiences of pioneers in early research on Internet routing, protocol conversion, and Software Defined Networking (SDN) has led me to realize that understanding and accurately describing the problem is more critical than merely focusing on the results. Only through an in-depth investigation of the problem itself can we lay a solid foundation for subsequent research and technological advancements. Additionally, learning from the experiences of predecessors and recognizing the challenges in research and practice can inspire us to adopt more comprehensive and innovative solutions when facing new problems.