Detail

Consistency without concurrency control

A CRDT is a data type whose operations commute when they are concurrent.
Replicas of a CRDT eventually converge without any complex concurrency con-
trol or the need for any centralised component. This makes CRDTs very ap-
pealing for managing data in large-scale environments, such as cloud computing
or web-based environments.

We have previously developed two non-trivial CRDTs: a shared edit buer,
Treedoc, and a graph structure, the multilog. This work allowed us to identify
some general properties for the design of CRDTs. The goal of this work is to generalise this approach to manage data in large-scale environments, by designing CRDTs for specic problems (e.g. replicated key-value store as used in Amazon Dynamo).
A CRDT is a data type whose operations commute when they are concurrent.
Replicas of a CRDT eventually converge without any complex concurrency con-
trol or the need for any centralised component. This makes CRDTs very ap-
pealing for managing data in large-scale environments, such as cloud computing
or web-based environments.

We have previously developed two non-trivial CRDTs: a shared edit buer,
Treedoc, and a graph structure, the multilog. This work allowed us to identify
some general properties for the design of CRDTs. The goal of this work is to generalise this approach to manage data in large-scale environments, by designing CRDTs for specic problems (e.g. replicated key-value store as used in Amazon Dynamo).