I attended HotEdge 2020 last week. When I submitted the paper in mid-February I had hardly any idea that the world was about to change so much in the coming weeks and months. Originally the conference was scheduled to take place in end-April 2020 in Santa Clara. Now it took place virtually on 25-26 June, 2020, right on the edge. :)
Edge is so much more: This is my first participation at the HotEdge workshop. Though in the past I have attended HotCloud and HotOS, I expected HotEdge to be something like those workshops. I was happy to learn and see many papers about infrastructure synthesis, infrastructure services (storage, networking, benchmarking, applications), building benchmarking, usecases in multiple domains (cellular networks, autonomous driving, smart home, space) and of course, security and trust. I presented work on Griffin, a shared state and data sharing service for collaborative applications at the edge. As someone who is trying to get started in this area, I found the workshop discussions quite useful. The key takeaway message for me was that the field is still in an early stage, and there are many fundamental problems that need solving before writing and deploying edge applications become as convenient as cloud applications today. Overall, it has been a really good experience.
On a personal level, I found the atmosphere at the workshop very welcoming and encouraging for everyone to join and participate. As Irfan Ahmad and Ming Zhao (co-chairs) proclaimed in the beginning of the workshop: “There are no “attendees” at HotEdge :-)… we are all workshop participants.” A big thanks to them, all the conference members, and the USENIX organization staff. I look forward to joining HotEdge 2021!
Experience during the workshop: To my surprise the online participation was quite smooth. No A/V issue for me (probably because we have been doing online meetings now for four months). The single track nature of the workshop was very helpful to keep the attention on a single ongoing activity. There were no multiple information channels (to decrease the cognitive load), just the main Hopin (the virtual platform used by HotEdge) window and a Google Doc on the side. The pre-recorded talks worked, and also helped keeping track of the time. The overall participation was high. In all activities there were 50-80 participants (numbers from my own highly scientific method of “periodic” peek at the numbers of participations). It never looked awkward and we never ran out of things to talk about or discuss. The majority of the participants were from Europe and the US, hence the 8am PST time slot to start the conference worked nicely.
I liked the interaction through a shared Google Doc. The doc was used to ask questions and get answers from the authors. The doc also served as a log for the worksop, summarizing the discussions on each paper, although there were some challenges (see more on this later).
There was no keynote speaker, but instead HotEdge had “fireside” chats with two guests, Delano Seymour (EDJX), and Subbu Varadarajan (Zycada). These chats were quite interesting. Irfan did an excellent job interviewing the guests. During the chat, the Google Doc was used to ask questions, which Irfan picked up to ask the guests directly. I thought this was quite nice, as I (and others) could participate in the discussion without interrupting the flow.
The networking event was enjoyable too. During the networking event, participants were paired randomly for 5 minutes to have a chit chat. I got paired with some known and unknown folks, which I enjoyed talking to. In one instance, I also got paired with my HotEdge paper’s co-author Lin Wang, which was funny (perhaps there are certain things computers are not good at!). We took this time to catch up on our impressions of the workshop so far.
What was challenging: There were a few things that were challenging for me. First, even though the networking event was nice, I missed corridor talks and bumping into people. Having 1-1 talks requires very active participation and energy, which was hard to do as the evening progressed here in Europe. Sometimes it took a while to do matches, while I waited nervously looking at my networking window. This way of interaction also precludes passive participation where I can just join in (standing on the side) an ongoing discussion and learn new things. I also found that five minutes was too little to discuss anything meaningful beyond basic introductory conversations even though I understand that the idea was to make an introduction and then follow up later. Travelling to a conference (often in nice locations) also offers a bit of a break from the daily routine of sitting in front of the computer at one’s office. The process of participating in HotEdge felt like joining yet another conference call in front of my computer. After a couple of hours of audio and video feeds, it was hard to continue to focus (breaks were not planned for HotEdge).
The formatting of the Google Doc was challenging. The doc itself scaled well, i.e., 80+ people could read and write at the same time. However, during a high activity session your cursor will be moving up and down randomly, and before you know it you would break a line for other participants. There was a lack of formatting (some color coding would have been nice). It would have been useful to have a fixed stable area for typing per user without messing up the content of others – may be an atomic transaction for writing and formatting in a shared doc. :)
Lastly, the poster sessions did not work for me. A few of them that I checked, I found myself in the middle of a slideshow presentation, just like another paper presentation, which was not very interesting or informative. It was hard to tune in. I missed looking at the one page visual of a poster. I think the general spontaneity of a group discussion (like during a poster session) is hard to capture online.
Overall: I enjoyed the workshop very much. For the next year, even if there is a physical conference I would still like some features of the online workshop to be included. Broadly speaking, I think we should offer online participation as a standard way of joining conferences besides physical attendance. To encourage participation, we can also consider reducing the participation fee for online participation. Online participation leaves more artifacts (vidoe, logs, Q/A, notes), is economical (time, money, energy required for participation), and greener to the environment (travelling).
PS~ We have 2 PhD position openings in the area of Edge Computing at VU Amsterdam. We are part of the CompSys group. If you are interested then drop me an email with your CV.