A Novel non Blocking Algorithm for Eventually Consistent Replicated Databases

Abstract

One of the challenges of cloud programming is to achieve the right balance between availability and consistency in a distributed database. Cloud computing environments, particularly cloud databases, are rapidly increasing in importance, acceptance and usage in major applications, which need the partition-tolerance and availability for scalability purposes, thus sacrificing the consistency side (CAP theorem). With this approach, use of paradigms such as eventual consistency became more widespread. In these environments, a large number of user’s access data stored in highly available storage systems. In this research one of the consistency model based on eventually consistent transaction using revision diagram for visibility and arbitration to manage the update in replica was discussed. A new algorithm was proposed that is sufficient for implementation of eventual consistency and how to use revision diagrams and fork-join data types to implement eventual consistency in a distributed system. It works as those eventually consistent transactions, where changes are immediately visible locally and eventually propagated to all other replicas in order to obtain better performance and availability. The new algorithm is validated by implementation of a pseudo code and using a case study.

Keywords: cloud computing, eventually consistent, replication