The 27th ACM Symposium on Operating Systems Principles (SOSP) was held in the beautiful Deerhurst Resort in Huntsville, Canada from October 27-30, 2019. Over 500 researchers from both academia and industry attended SOSP 2019. The conference was organized by Tim Brecht and Carey Williamson. The PC chairs Remzi Arpaci-Dusseau and Yuanyuan Zhou put together a great program consisting of 38 papers. Authors presented innovative ideas on machine learning, security, privacy, verification, and more.
There were four workshops organized at SOSP this year. The AI Systems workshop is held at a systems and ML venue (like NeurIPS) every year, and aims to foster the MLSys community. The Workshop on Programming Languages and Operating Systems (PLOS) aims to bring together researchers from the OS and PL communities to work on problems of mutual interest. The Workshop on System Software for Trusted Execution (SysTEX) focused on work that aims to make trusted execution environments like Intel SGX more usable. Finally, the Diversity workshop provides a forum for women and other under-represented minorities.
The workshops are an important tradition in SOSP, fostering cross-discipline collaboration and nurturing new fields. We hope this continues in future editions of the conference!
The main program was split across 11 sessions on three days. There were a number of major themes, such as machine learning (five papers), verification (four papers), bug finding (four papers), distributed systems (four papers), and storage (four papers). Other focus areas included cryptocurrencies (two papers) and graph engines (two papers). While the program is not determined keeping categories in mind, this gives a good idea of what folks in the community are working on, and what the community considers important.
A number of papers from SOSP were featured in the Morning Paper blog, which discusses an interesting CS research paper every day:
- An analysis of performance evolution of Linux’s core operations
- File systems unfit as distributed storage backends: lessons from ten years of Ceph evolution
- The inflection point hypothesis: a principled approach to finding the root cause of a failure
- Snap: a microkernel approach to host networking
- Scaling symbolic evaluation for automated verification of systems code with Serval
- Taiji: managing global user traffic for large-scale Internet services at the edge
Authors presented posters for all accepted posters on the evening of the first day. Conference attendees enjoyed snacks and drinks in a relaxed atmosphere. This allowed the audience to engage more deeply with the content and ask follow-up questions to the authors.
Malte Schwarzkopf (Brown) and Natacha Crooks (Berkeley) put together a Mentoring scheme for the first time at SOSP. Students were matched to a researcher or faculty member. The mentoring scheme aims to make it easier for students to meet and interact with the more senior members of the community. We have had first hand from several students that the mentoring scheme made their conference experience much better! Kudos to Malte and Natacha, and we hope this is continued at the next SOSP.
Baris Kasikci (Michigan), Supreeth Shastri (UT Austin), and Vijay Chidambaram (UT Austin) put together Artifact Evaluation for the first time at SOSP. 23 papers out of the 38 accepted papers opted to undergo Artifact Evaluation. Out of the 23, 22 earned at least the Artifact Functional Badge, and 11 earned all three badges offered: Artifact Functional, Artifact Available, and Results Reproduced. The Artifact Evaluation was only possible due to the efforts of the Artifact Evaluation Committee. More details are available here. We hope Artifact Evaluation happens at future instances of the conference as well.
At the dinner banquet on the second day, various awards were given out. This included the Best Paper Awards, Test of Time Awards, the Mark Weiser Award, and the Dennis M. Ritchie Doctoral Dissertation Award. The winners of the Student Research Competition were also announced. More details on the awards and the winners are available here.
A couple of topics inspired interesting and spirited discussions at the dinner banquet. The first one was by SIGOPS Chair Shan Lu (Chicago), about increasing the acceptance rate at SOSP. SOSP 2019 had a 13.7% acceptance rate, significantly lower than top conferences in other areas (which tend to be between 16-20%). Shan made the case that the number of papers submitted to SOSP has gone up over the years, but the number of papers accepted has not. The second topic was by John Ousterhout (Stanford), who made the case for removing page limits in camera-ready versions of accepted papers. After some discussion, the consensus among participants seemed to be towards giving paper shepherds more flexibility in awarding extra pages to an accepted paper. Both topics generated a lot of discussion: we will be providing more details in this blog later!