https://www.dagstuhl.de/21462

November 14 – 17 , 2021, Dagstuhl Seminar 21462

Foundations of Persistent Programming

Organizers

Hans-J. Boehm (Google – Mountain View, US)
Ori Lahav (Tel Aviv University, IL)
Azalea Raad (Imperial College London, GB)

For support, please contact

Dagstuhl Service Team

Documents

Dagstuhl Report, Volume 11, Issue 10 Dagstuhl Report
Aims & Scope
List of Participants
Dagstuhl Seminar Schedule [pdf]

Summary

We brought together 15 in-person attendees at Schloss Dagstuhl, with a roughly equal number of remote attendees. Remote attendance was challenging, particularly for attendees from very different time-zones. It nonetheless provided the opportunity for us to hear from a wider selection of participants.

We decided up-front not to try to cater the schedule to remote participation. Given the number of time zones covered by the participants, we continue to believe that, although it clearly had adverse impacts, it was the right decision.

We had a number of remote presentations that included interactions with the speakers. Otherwise the discussion tended to happen mostly among the in-person participants, in spite of the excellent AV systems at Dagstuhl. Many remote participants were limited in attendance due to time-zone issues.

Our area is perhaps unique, in that it includes deep theoretical work, but is also very dependent on technological developments. Accordingly, the participants of the seminar were from a spectrum of topics ranging from theory of distributed systems to hardware specification and design. The seminar gave many of us the opportunity to catch up on both theory and practice, including input from some participants with more direct insights into industrial developments.

We began the seminar by reviewing some of the underlying assumptions that were made by prior work in this field, often without certainty about their correctness. We were actually able to get much more shared clarity on a few of these as a result of audience discussion during the seminar. Some of us learned that non-volatile caches are being publicly discussed by Intel, and that there also is similar agreement that memory encryption, to restore volatility when needed, is desirable. We also learned that writes to the same cache line are not just believed by software researchers to reach memory in the correct order, but at least one hardware vendor also agrees. Though this last fact is rather obscure, it is important for some NVM algorithms, and not normally reflected in hardware manuals.

The rest of the seminar consisted of talks and group discussions. Three talks were longer overview talks on different aspects (Michael Scott on buffered persistency, Parosh Aziz Abdulla on verification, and Erez Petrank on persistent lock-free data structures). We did not feel the need for smaller break-out sessions, since the in-person group was quite small, largely due to our timing with respect to Covid waves. Much of the benefit here appears to have been in listening to discussions that often were either significantly more theoretical or significantly more practical than our own research.

NVM programming is both complicated by, and often synergistic with concurrent programming. Much of our focus was on the interaction between the two. Due to these close interactions, we asked several speakers to talk about concurrency issues that seemed particularly relevant (e.g., Peter Sewell on Armv8-A virtual memory model, Mark Batty on novel solution to the ``thin air problem'', and Paul McKenney on weak memory schemes used in the Linux kernel).

Several talks raised questions on the foundations of the field, such as what are the hardware-supplied programming models, or what it means for a persistent program to be correct. It is entirely possible that most future NVM programmers will be more concerned with something like the persistent transactions discussed in Michael Bond's talk. However, given the relative immaturity of the area and foundational uncertainty, the emphasis seemed appropriate.

Summary text license
  Creative Commons BY 4.0
  Hans-J. Boehm, Ori Lahav, and Azalea Raad

Related Dagstuhl Seminar

Classification

  • Distributed / Parallel / And Cluster Computing
  • Programming Languages

Keywords

  • Non-volatile-memory
  • Persistency
  • Semantics
  • Weak memory models
  • Concurrency

Documentation

In the series Dagstuhl Reports each Dagstuhl Seminar and Dagstuhl Perspectives Workshop is documented. The seminar organizers, in cooperation with the collector, prepare a report that includes contributions from the participants' talks together with a summary of the seminar.

 

Download overview leaflet (PDF).

Dagstuhl's Impact

Please inform us when a publication was published as a result from your seminar. These publications are listed in the category Dagstuhl's Impact and are presented on a special shelf on the ground floor of the library.

Publications

Furthermore, a comprehensive peer-reviewed collection of research papers can be published in the series Dagstuhl Follow-Ups.