Simizer: evaluating consistency trade offs through simulation

Abstract

Eventually consistent data stores are a widely used solution to the scalability issues arising for modern applications and web based services. Due to the scale of such services, it is necessary to be able to simulate such infrastructures and application protocols in order to assess the correctness and performance of the system over realistic workloads at development time. This paper describes Simizer, a new cloud computing platform simulator focusing on distributed services protocol performance evaluation. Usability of this simulator is shown through a sample implementation of an optimistic replication policy for a distributed storage service.