1: \begin{abstract}

2: %Eventual consistency is a relaxed consistency model used in large-scale distributed systems that seek better availability when consistency can be delayed. 

3: 

4: CRDTs are distributed data types that make eventual consistency of a

5: distributed object possible and non ad-hoc. Specifically, state-based CRDTs

6: ensure convergence through disseminating the entire state, that may be large,

7: and merging it to other replicas; whereas operation-based CRDTs disseminate

8: operations (i.e., small states) assuming an exactly-once reliable

9: dissemination layer.

10: %with an idempotent, associative, and commutative join operation,

11: %ensuring convergence. 

12: %This imposes a large communication overhead as

13: %the state size becomes larger. 

14: We introduce \emph{Delta State Conflict-Free Replicated Data Types} (\dsrdt)

15: that can achieve the best of both worlds: small messages with an incremental

16: nature, as in operation-based CRDTs, disseminated over unreliable

17: communication channels, as in traditional state-based CRDTs.  This is achieved

18: by defining \emph{$\delta$-mutators} to return a \emph{delta-state}, typically

19: with a much smaller size than the full state, that to be joined with both

20: local and remote states.

21: %(after being shipped). 

22: We introduce the \dsrdt framework, and we explain it through establishing

23: a correspondence to current state-based CRDTs.

24: In addition, we present an anti-entropy algorithm for eventual convergence,

25: and another one that ensures causal consistency. Finally, we introduce several

26: \dsrdt specifications of both well-known replicated datatypes and novel datatypes,

27: including a generic map composition.

28: 

29: %\begin{IEEEkeywords}

30: %Distributed systems, eventual consistency, state-based, CRDT.

31: %\end{IEEEkeywords}

32: \end{abstract}

33: