Evaluation of the CAP Properties on Amazon SimpleDB and Windows Azure Table Storage

Abstract

The advent of Web 2.0 and the diffusion of cloud computing technologies recently moved the focus of industry and academia to new ways to store information. Web 2.0 applications require new solutions that allow to efficiently store and compute huge amounts of data, providing high performance, high scalability and availability at low costs. Even though relational databases are mature and offer powerful tools to manage data, their scalability limits have brought some of the largest IT companies to adopt less mature solutions, known as NoSQL databases. In 2000, a conjecture, formalized afterwards as the CAP theorem, concerning the Consistency (C), Availability (A) and Partition tolerance (P) of distributed systems was formulated. In this paper we propose a methodology based on the CAP theorem and a framework for the assessment of the CAP properties of cloud NoSQL systems. Furthermore, we provide the results of an extensive set of experiments we performed on Amazon SimpleDB and Windows Azure Table Storage for comparing quantitatively Availability and Consistency of such systems.