- Topic registration: link will be emailed to you on Mar 17 (Fri) 12:00PM noon time (first-come-first-serve basis).
- Submission: May 12 (Fri) 11:59pm. Submit the presentation slides and link of a 20-min recorded video on Blackboard.
Project Guidelines for IERG5090
To design your own project topic, you are strongly encouraged to browse through a group of related papers (and the papers they refer to) which appeared in the following conferences in the past few years: ACM Sigcomm, USENIX NSDI, ACM IMC, ACM Mobicom, ACM Sigmetrics, USENIX ATC, EuroSys, USENIX OSDI, ACM SOSP.
For the topic of your choice, provide a critical analysis and comparison of the different solution approaches you found (from the literature) for the topic. One suggestion is to start by giving an overview of the topic and identify the key problems related to the topic. You can then organize the list of papers you read and show how they address (are related to) these key problems. You then proceed to discuss, for each (or a selected subset of) paper (or solution) you read, its key ideas/ innovations, assumptions, analytical/ algorithmic techniques, major results, their significance, and potential level of impact with justifications. For example, why you think the findings of the paper is important and in what way can it help other researchers to advance the field further. You may also propose possible extensions of the paper, e.g. by examining carefully the assumptions of the paper and discuss how the problem/ result will change if those assumptions change or get modified.
Your presentation should go beyond merely summarizing individual papers: you should compare and contrast the results/ findings across those papers and be able to give a status of the state-of-the-art in the area, and remaining challenges (i.e. new problems/ extensions) and opportunities waiting to be addressed. In general, your presentation should not discuss individual papers (solutions) as isolated objects but show how those solutions compare with each other and/or how they complement each other. The ultimate goal is to provide a big picture on the state-of-the-art (industrial and/or research) landscape, available solutions and their competitive analysis associated with your topic. Original results will be a big plus but not required. Each student should submit the presentation slides and link of a 20-min recorded video on Blackboard.
NO late submissions will be accepted as we need to grade your projects and submit the final marks shortly after the deadline.
Here are some possible/suggested project topics for your consideration:
- In-Network Machine Learning, Network Meets AI & ML (e.g. the Taurus work by Prof. Kunle Olukotun of Stanford/Samba Nova ; Nick Feamster of U of Chicago ; ACM Sigcomm NetAI workshops)
- SmartNICs (e.g. from Ming Liu et al of UW-Madison/VMware/U of Washington; Alex C. Snoeren ofUCSD; Tonic by Jennifer Rexford et al)
- Network Verification (e.g. Mina Tahmasbi Arashloo of U of Waterloo/Cornell/Princeton, George Varghese of UCLA, Nate Foster of Cornell ; the Petr4 work)
- P4 and its applications/ Programmable Data Plane (e.g. Minlan Yu of Harvard on SDN switch validation ; P4 Campus work from Jennifer Rexford of Princeton)
- Network Telemetry, Programmable Network Measurement Architecture / Network Telemetry (e.g. Minlan Yu of Harvard, OmniMon from Patrick Lee of CUHK; Evolvable Network Telemetry at Facebook, NSDI 2022, Sonata from Princeton)
- Data-center Networking/ Network Management (e.g. Mohammad Alizadeh of MIT, Mosharaf Chowdhury of U of Michigan, Van Jacobson/David Wetherall of Google)
- RDMA for Data-center Networks (e.g. Sigcomm 2018 paper by Arvind Krishnamurthy, Sylvia Ratnasamy, Scott Shenker, references therein and follow-up works ; 1RMA work from Aditya Akella et al ; Collie from NSDI 2022).
- Advances in Congestion Control (e.g. the Pantheon Project from Keith Winstein of Stanford, HPCC from Minlan Yu of Harvard, Hari Balakrishnan of MIT, Mohammad Alizadeh of MIT)
- Evolution and Comparison of the Cloud Networking Infrastructure from (possibly a subset of the following cloud-service providers) Google, Facebook/Meta, Microsoft Azure, Alibaba, Byte Dance etc, e.g. based on Sigcomm publications such as those mentioned in [1, 2, 3]
You can also refer to the recent research work of the following networking researchers and then propose a related topic about one of their recent projects, references therein and follow-up works by other groups.
For the case of a self-proposed topic by the student, please send an email to Prof. Wing C. Lau to seek approval in advance.
- Aditya Akella (U of Texas/ U of Wisconsin)
- Mohammad Alizadeh (MIT)
- Zakir Durumeric (Stanford)
- Nick Feamster (U of Chicago)
- Nate Foster (Cornell)
- Brighten Godfrey (UIUC)
- Dongsu Han (KAIST)
- Kyle Jamieson (Princeton)
- Xin Jin (Peking University/ Johns Hopkins)
- Ethan Katz-Bassett (Columbia)
- Praveen Kumar (Google/ Cornell)
- Ming Liu (U of Wisconsin)
- Vincent Liu (of UPenn)
- Ratul Mahajan (U of Washington)
- Nick McKeown (Intel/ Stanford)
- Amy Ousterhout (UCSD)
- Simon Peter (U of Washington/ U of Texas)
- Jennifer Rexford (Princeton)
- Vyas Sekar (CMU)
- Alex Snoeren (UCSD)
- Mina Tahmasbi Arashloo (U of Waterloo)
- Renata Teixeira (Inria)
- Amin Vahdat (Google)
- Laurent Vanbever (ETH)
- George Varghese (UCLA)
- Keith Winstein (Stanford)
- Minlan Yu (Harvard/Yale/USC/Princeton)