https://www.dagstuhl.de/18091

February 25 – March 2 , 2018, Dagstuhl Seminar 18091

Data Consistency in Distributed Systems: Algorithms, Programs, and Databases

Organizers

Annette Bieniusa (TU Kaiserslautern, DE)
Alexey Gotsman (IMDEA Software – Madrid, ES)
Bettina Kemme (McGill University – Montreal, CA)
Marc Shapiro (University Pierre & Marie Curie – Paris, FR)

For support, please contact

Dagstuhl Service Team

Documents

Dagstuhl Report, Volume 8, Issue 2 Dagstuhl Report
Aims & Scope
List of Participants
Shared Documents
Dagstuhl Seminar Schedule [pdf] (Update here)

(Use seminar number and access code to log in)

Summary

Large-scale distributed systems have become ubiquitous, and there are a variety of options to develop, deploy, and operate such applications. Typically, this type of application is data-centric: it retrieves, stores, modifies, forwards, and processes data from different sources. However, guaranteeing availability, preventing data loss, and providing efficient storage solutions are still major challenges that a growing number of programmers are facing when developing large-scale distributed systems. In our seminar, we brought together academic and industrial researchers and practitioners to discuss the status quo of data consistency in distributed systems. As result of talks and discussions, we identified several topics of interest that can be grouped into the following four areas.

Theoretical foundations: The seminar included a tutorial on specification of consistency guarantees provided by distributed systems and talks on comparing different styles of specification and expressing replicated data type semantics in Datalog. Different specification styles are suitable for different purposes and more work is needed to identify the most appropriate ones. The seminar also included talks on formally reasoning about which consistency levels are enough to satisfy correctness properties of applications. The talks demonstrated that formal verification is a promising approach to cope with the challenge of selecting appropriate consistency levels.

Distributed systems and database technologies: With the growing number of replicated data stores, the two fields of distributed systems and databases are moving closer together. The communities should be made more aware of each others results. A common concern in agreement, i.e., ensuring that database copies are updated correctly. Traditionally, the distributed systems community has based many of their approaches on classical consensus algorithms or looked at weaker consistency models. In contrast, database systems focused most work on 2-phase commit protocols and eager update protocols. At the same time, the database community also considered other ACID aspects that required to combine commit protocols with concurrency control protocols and recovery schemes. In the last decade however, and in particular with practical implementations of the Paxos consensus algorithms, and the use of file replication in storage systems for availability, work of the two communities has come closer together. A challenge in this context is that work that emerges from the different communities still makes slightly different assumptions about failure and correctness models. They can often be quite subtle so that the differences are not obvious, even to the experts. And they can lead to very different approaches to find solutions. Bridging this gap in terms of understanding each other, and the implications of correctness and failure models remains a challenging task. As an example, the separation of the concepts of atomicity, isolation and durability in the database world offers many opportunities for optimization, but includes extra complexity when analyzing which algorithms are appropriate in which situations.

Conflict-handling in highly-scalable systems: In the last years, conflict-free replicated data types (CRDTs) have been adopted by an ever-growing number of products and companies to deal with high-availability requirements under concurrent modifications of data. Recent advances in related techniques for collaborative editing might make it possible that hundreds of people work together on a shared document or data item with limited performance impact. Several talks presented programming guidelines, static analyses, and related tools for safe usage of CRDTs in situations where eventual consistency is not enough to maintain application invariants.

Programming models for distributed systems: Micro-services have become a standard approach for constructing large-scale distributed systems, though microservice composition and scalability raises a lot of questions. Some presentations discussed current work on actor-based and data-flow programming. Design for testability and test frameworks are crucial for providing reliable services, but they currently require a lot of experience as of today. We believe that future progress on programming models and new results in theoretical foundations will help to simplify this challenging task and support programmers in building safe systems.

License
  Creative Commons BY 3.0 Unported license
  Annette Bieniusa, Alexey Gotsman, Bettina Kemme, and Marc Shapiro

Related Dagstuhl Seminar

Classification

  • Data Bases / Information Retrieval
  • Programming Languages / Compiler
  • Verification / Logic

Keywords

  • Distributed Computing
  • Consistency
  • Replication
  • Partitioning
  • Transactions

Book exhibition

Books from the participants of the current Seminar 

Book exhibition in the library, ground floor, during the seminar week.

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).

Publications

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

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.

NSF young researcher support