PODC Archives

ACM PODC Participants List


Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Marc Shapiro -- at work <[log in to unmask]>
Reply To:
Marc Shapiro -- at work <[log in to unmask]>
Mon, 14 Mar 2011 09:20:03 +0100
text/plain (20 lines)
"Principled approaches to eventual consistency "
Post-doc offer.  Apply preferably by 15 March 2011.


Data replication is a fundamental mechanism of distributed systems, for performance, availability and fault tolerance. To sidestep the incompatibility between strong consistency and fault-tolerance, an attractive approach is eventual consistency (EC). EC allows a replica to update without synchronising; concurrent updates are reconciled in the background. However, existing EC algorithms are often ad-hoc and/or unpredictable. Conflict resolution requires synchronisation and is complex and error-prone.

Regal has previously proposed a principled approach to EC, the action-constraint formalism, implemented in the Telex platform. A recent insight is that it is possible to design non-trivial and generally useful data types whose concurrent operations commute (we call them CRDTs). This ensures EC in a simple and sound way, requiring no synchronisation. Our challenges so far have been to design a library of useful and interesting CRDTs, and to characterise sufficient conditions for consistency. Many research questions remain open: how do we build interesting, large-scale applications using CRDTs? Is the CRDT approach sufficient, or is some synchronisation always necessary; and if so, how much, and how do we fit the two together? What are the theoretical and practical limits of CRDTs? What complexity bounds and real performance? How do CRDTs relate to commutativity theory, monotonic languages, self-stabilisation, or the consensus hierarchy?

We wish to recruit a post-doc researcher interested in pursuing this topic. Candidates should have an excellent research record, with previous experience in distributed systems, distributed algorithms, distributed databases, consistency, or large-scale distributed applications. Candidates should be motivated by the theoretical and practical issues of eventual consistency.

Regal is a joint group of INRIA and LIP6, located in Paris. Our area of research is computing systems (interpreted broadly). We focus mainly on large-scale distributed systems, which include grids, cloud and P2P systems, for applications such as social networks or on-line games. Our research interests cover the interaction between theory and practice. We also study software for multi-core computers, application virtual machines, and development tools for operating systems. Our recent publication venues include JPDC, IPDPS, TAAS, SRDS, ICDCS, DISC, SSS, SPA, ASPLOS, VEE, TSE, etc.

Contact and application information: 
Application form:
Post-doc platform of INRIA: