Members
Overall Objectives
Research Program
Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography
XML PDF e-pub
PDF e-Pub


Section: Software and Platforms

SwiftCloud

Participants : Mahsa Najafzadeh, Burcu Külahçioglu Özkan, Marc Shapiro [correspondent] , Marek Zawirski.

Cloud computing infrastructures improve latency and provide high availability by geo-replicating data at different locations across the world. This improves user experience, which is important for services such as social networks, online shops and games. Nevertheless, the distance to the closest data centre is still far from optimal for many users.

SwiftCloud is the first system to bring geo-replication all the way to the client machine, in order to provide the best possible latency and availability. This raises two main challenges. One, how to provide, efficiently, convenient programming guarantees, including access to consistent data in read-write transactions, and ensuring session guarantees. Two, to continue providing these guarantees, despite failures that require a client to switch to a different data centre.

Our research report [61] presents the design of SwiftCloud and the algorithms it uses to achieve the desired properties, while aiming for efficiency and for scalability to large numbers of clients. Our evaluation confirms that its programming model is practical, and that its performance and fault tolerance objectives are met.

SwiftCloud is supported by the ConcoRDanT ANR project (Section  8.1.6 ), by a Google European Doctoral Fellowship, and by the new FP7 grant SyncFree (Section  8.2.1.1 ).

The code is freely available on http://gforge.inria.fr/ under a BSD license.