Background: The ChinaSys community just held a workshop in June 2024 in Hangzhou. The workshop added a Best Paper session (3 papers) and a Lightning Talk + Poster interactive session (54 papers), enabling participants to engage in extensive discussions on various aspects of systems such as architecture, OS, cloud computing, storage systems, machine learning systems, graph computing and so on. The workshop received a total of 88 submissions, covering top-tier conferences including OSDI, ASPLOS, FAST, EuroSys, ISCA, HPCA, SC, PPoPP, ATC, VLDB, DAC, NSDI, among many others. The total number of attendees surpassed 400, reaching a record high. Additionally, the workshop implemented live online streaming, garnering widespread attention with over 130,000 views of the video streams and nearly 150,000 visits to online photo broadcasts.
This article will share my experience to attend the 26th ChinaSys workshop (ChinaSys 2024 Hangzhou), and my thoughts about how to build a better ChinaSys workshop.
I used to be the one (as a PhD student) who benefited from ChinaSys very much. ChinaSys has always been an amazing systems workshop to me. It has accompanied my research growth, from when I was a novice, to giving talks on stage, and now actively thinking about how to improve it.
Early in 2022, I had a discussion with Professor Yubin Xia from SJTU. He mentioned that the total number of folks (faculty and students included) in the ChinaSys community was fewer than a thousand people. I replied that for these folks including us, it was our golden age (meaning that doing a little could be seen soon in the community), but it has not reached the golden age of Chinese system research.
My thoughts as follows:
Workshop duration: Since most faculty and students need to work on weekdays, the 2-day workshop duration on the weekend is decided to be a hard limit. However, the paper submissions are exploding, why are we stuck in the single-track convention? In the early days, the number of people in both China Architectures and Systems was not large, so we united. Now, as the work in the field is developing so rapidly, should Chinese Architectures and Systems be divided? I think not; otherwise, software-hardware co-design would never be achieved (the silos situation between different fields should be avoided). I hope that the number of tracks can be appropriately increased. Take USENIX ATC/OSDI as an example, they now look like a double-track conference (thanks to the 2024 USENIX organizers who deliberately staggered sessions with similar topics so that everyone can switch between ATC/OSDI). So, whether to add more tracks is a question.
Lightning talks: This year’s ChinaSys has a very big improvement, that is, the introduction of lightning talks. Thanks to the efforts of Dr. Shuang Chen and Xingda Wei! Personally, I think lightning talks are suitable to be released as pre-workshop teasing videos, and this session can be omitted during the conference. Of course, Dr. Shuang also reminded me that lightning talks give presenters opportunities to be exposed on stage, and hence it is also worth keeping.
Mentorship: During the dinner party, everyone tends to gather around acquaintances, but meeting more people from different backgrounds can be a better choice. Mentorship can be an exciting moment during the party (so is Ask Me Anything). During registration, everyone can choose their favorite topics and the reviewers they want to meet as mentors. We evenly distribute the mentors to different dinner tables. For mentors with too many mentees, reshuffling can be done halfway through.
Posters: Also, can the poster session be integrated with the dinner party, supplemented by robots delivering pastries automatically? As students, they certainly have no resistance to food, and even less resistance to “spiritual food” such as the knowledge and experience of their peers. We should strive to expand the time of the poster session, allowing students to fully immerse themselves in a lively discussion vibe, and complete a wonderful brainstorming. It has to be said that the speakers of the poster session do not have to face the nervousness of being stared at by countless eyes on the podium. They can be more excited in a more relaxed situation and make more good friends. Even introverted people who are not good at speaking can “steal” a lot of experience by shuttling between different posters. Therefore, sufficient time, chill vibe, and nice food should be provided for the poster session! Personally, it can be held together with the dinner party. Everyone can get up and walk around after eating for a while.
Industry Session: Back to the industrial orals of ChinaSys, we definitely welcome recruitment from the industry. Academia is responsible for providing problem-solving talents, and the industry is responsible for providing real problems and attractive salaries, hopefully accomplishing a win-win! As more and more outstanding students from our academia go to the industry, let them express the interesting problems in the language that students in the ivory tower can understand, and express the gap between the two and thoughts on how to enhance influence. For example, what skillsets do the industries want? What research is useful and can be impactful? What models can allow different universities to participate, replicate, and thereby give full play to everyone’s enthusiasm and creativity?
Orals: It has to be said that the levels of oral presentations vary. The reason is that there are too few good templates to study and a lack of training. I personally love “How to Give a Bad Talk” by David Patterson. Most reports are like giving group meeting talks rather than TED-like stories with fascinating plots. The principle I follow is the cache principle (credit to Dr. Cheng Tan):
The audience’s cache is small, so don’t give too many concepts; there should be locality, that is, what is said 3 pages ago must serve the later part and not be too far apart.
A good talk requires preparation for more than 1 month and needs to be presented to colleagues from different fields for continuous revision. Good slides need to be self-explained by only looking at the text and pictures. Don’t be too greedy in slides; clearly explain the problem and results, and just give an insight into the solution (everyone can establish the relationship between solutions and results through logic). Let them read your paper for the remaining details.
My suggestion is that If the paper was just accepted and the talk slide has not been prepared for more than 2 months, then leave it for a while and look back to check whether the logic can be improved (similar to code refactoring). If yes, you had better wait until the next ChinaSys (ChinaSys is held twice a year).
Submissions: I like the 2-page extended abstract (initiated by ASPLOS). How to identify interesting and fun stories quickly and accurately? We may submit a 2-page long abstract (which is actually a highly structured Intro), a 4-minute lightning talk, and select the appropriate work based on the presentation quality. I believe that students who attend the conference would like to hear interesting and fascinating stories and learn conclusions that we haven’t heard in textbooks and blogs.
Participation: How to participate in the conference also requires guidance. There should be something like a cheat sheet, similar to the “SIGPLAN Empirical Evaluation Checklist” for AEC members. When students go to the conference, how do they introduce themselves? Whom do they want to meet and which peers do they want to know? What questions to ask (including silly and good questions)? Which talks to listen to? Can they find a suitable person for future jobs? Remember, your time is as precious as that of the big shots. How to cherish your time and make the most of this trip?
Misc: In previous years, there were also forums for experience sharing by outstanding doctoral graduates (such as “The Story Behind Scientific Research” in ChinaSys 2020 Chongqing), all full of dry goods. I miss them. I believe that the old comrades of ChinaSys don’t want to always miss the old golden ages but want to create a nova platinum era for the new generation. Boyu Tian from Tsinghua University suggested that there could be some debates and organize students and teachers/industry peers to battle each other. I believe that students with debate experience can handle the heat well, being both intense and decent.
ChinaSys has been for more than 13 years and now reached a new era. It has well served the Systems community and can be even better. As the seeds of the spark that starts a prairie fire, I believe ChinaSys can play this role well.
About the author: Mingyu Li is an Associate Professor at Key Laboratory of Systems Software, Institute of Software, Chinese Academy of Sciences, working on operating systems and confidential computing.
Disclaimer: Please note that any views or opinions represented in this blog are personal, belong solely to the blog author and do not represent those of ACM SIGOPS. We welcome any constructive criticism.
Editor: Dong Du (Shanghai Jiao Tong University) and Tianyin Xu (University of Illinois at Urbana-Champaign)