Project Team Regal

Members
Overall Objectives
Scientific Foundations
Application Domains
Software
New Results
Partnerships and Cooperations
Dissemination
Bibliography
PDF e-pub XML


Section: Partnerships and Cooperations

European Initiatives

Google European Doctoral Fellowship “A principled approach to eventual consistency based on CRDTs

Cloud computing systems suffer from a fundamental tension between scalability and data consistency. Avoiding the synchronisation bottleneck requires highly skilled programmers, makes applications complex and brittle, and is error-prone. The Commutative Replicated Data Type (CRDT) approach, based on commutativity, is a simple and principled solution to this conundrum; however, only a handful of CRDTs are known, and CRDTs are not a universal solution. This PhD research aims to expand our knowledge of CRDTs, to design and implement a re-usable library of composable CRDTs, to maintain study techniques for maintaining strong invariants above CRDTs, and to experiment with CRDTs in applications. We are hopeful that significant distributed applications can be designed using our techniques, which would radically simplify the design of cloud software, reconciling scalability and consistency. This Google European Doctoral Fellowship is awarded to Marek Zawirski, advised by Marc Shapiro. This award includes a grant of 41 000 euros yearly over three years starting September 2010.

FTH-GRID - (2009–2011)

Members:

Université de Lisbonne (LASIGE), LIP6 (Regal)

Funding:

Egide

Objectives:

FTH-Grid, Fault-Tolerant Hierarchical Grid Scheduling, is a cooperation project between the Laboratoire d'Informatique de Paris 6 (LIP6/CNRS, France) and the Large-Scale Informatics Systems Laboratory (LASIGE/FCUL, Portugal).

Its goal is to foster scientific research collaboration between the two research teams. The project aims at rendering Map Reduce on top of Grid tolerant to byzantine failure. Map Reduce is a programming model for large-scale data-parallel applications whose implementation is based on master-slave scheduling of bag-of-tasks. MapReduce breaks a computation into small tasks that run in parallel on different machines, scaling easily to several cluster. The core research activities of the project consist mainly in extending the execution and programming model to make Byzantine fault-tolerant MapReduce applications. The project was extended for another year, after a results assessment by Egide.