The EuroSys 2020 Online Conference

The 15th European Conference on Computer Systems (EuroSys’20) was organized as a virtual (online) conference on April 27- 30, 2020. The main EuroSys’20 track took place April 28-30, 2020, preceded by five workshops (EdgeSys’20, EuroDW’20, EuroSec’20, PaPoC’20, SPMA’20) on April 27, 2020. The decision to hold a virtual (online) conference was taken in early April 2020, after consultations with the EuroSys community and internal discussions about potential options, eventually allowing about three weeks for the organization.

A recent report offers an account on the decision to go virtual, preparations to organize the online event, details on the execution, and insight into how well the event went based on survey feedback from 100 of the EuroSys’20 attendees. Overall, 72% of respondents found virtual EuroSys’20 to be better or much better than what they expected a virtual conference to be like, and 85% of respondents were satisfied or very satisfied with the organization of virtual EuroSys’20. Nearly all respondents agreed that augmenting EuroSys conferences with online features would be useful even for physical conferences in the future.

Key insights and lessons learned

EuroSys’20 synchronous Zoom sessions were modeled after the typical sequence of events at a physical conference, except that talks were shorter. We asked authors to provide two versions of each talk: a short (3-5 minute) version to be streamed at the conference, and a longer (10-12 minute) talk to be made available in advance so that attendees could watch in preparation of the actual talk. While this put an extra burden on authors, it proved an important element of success as it allowed us to experiment with talk durations and adapt in real-time, which we eventually did.

A dense program (43 talks in 2.25 days) along with multiple interactions and information channels, created at times a sense of information overload, especially for remote attendees, which are still immersed in everyday life (home, office). Reducing the duration of talks by streaming shorter videos (3 mins) did not seem to provide adequate technical detail to feed a live Q&A session.  Longer videos (10-12min) worked better, as was also noted in the responses of surveyed participants.

Based on the expectation that more online features would increase the level of participation, EuroSys opted to augment synchronous Zoom video conferencing with two asynchronous interaction platforms (Slack and Discord). While this allowed EuroSys’20 to be a live experimentation platform, which was well received by many, we would have preferred to use a single platform. As Discord could implement all asynchronous Slack functionality that we wanted and provide more in the direction of synchronous interactions (hosted virtual meetings, Hallway track), in retrospect we would have chosen Discord as the single platform.

Organizing a virtual conference requires significant team building and training, and new processes involving several new technologies. We found that the Zoom master (the EuroSys’20 role tasked with training other Zoom roles (hosts), creating play sequences, developing the registration system, managing registrations, and communicating with participants via mailing lists, before and during the conference) proved to be critical. Although the master could assign sub-tasks to a team of volunteers, in practice their tasks were not always parallelizable. In retrospect, we would have opted to split the Zoom master role to more than one person.

Training is key to success and it should start early. In EuroSys’20, we embarked upon training early in the (time-constrained) preparation process. Our Zoom master led the training of Zoom co-hosts so that they could serve as backups, and also as hosts of parallel workshop sessions. The Zoom master also trained Session chairs to be able to operate as co-hosts during sessions. Zoom co-hosts spun off and scheduled training tasks with workshop organizers. The plan was also to invite speakers for dry runs; however, this was not possible within the limited time available. In retrospect, it seems that it was not necessary either: speakers seemed to find it easy to join as panelists and participate in Zoom sessions for Q&A, so such training may in fact be an optional step. What helped significantly during training was the production of textual user guides for session chairs and presenters.

Effective coordination between organizers is another key to a successful virtual conference. To facilitate it, we set up private chat channels for easy communication and coordination between session chairs and presenters for each session, organizers (general chairs, PC chairs) and session chairs, local organizing team (general chairs & Zoom/Slack/Discord hosts), organizers and sponsors, and organizers and sponsor chairs. These private channels proved important for the smooth operation of the conference. For instance, we decided to change the length of videos from short to long versions at the end of session 1; this change was agreed upon and communicated among all organizers. It was applied without disruption to the program.

It is important for a virtual conference to engage with sponsors. While not addressed in detail in our report, the organization model put in place many opportunities for sponsors to interact with attendees (on Zoom and through Slack/Discord channels and Discord meeting rooms). We found that it is important to engage sponsors early on, to communicate opportunities, and to give them enough lead time to assign representatives to participate in activities (Zoom and channels) and to provide the necessary content.

More details and further insight on the EuroSys’20 experience are documented in the full report. Overall, we believe that the EuroSys’20 experience led to important insight into how to organize an effective virtual conference and in understanding the challenges that remain ahead. We hope that this experience will improve the organization of virtual and physical conferences in the future.