Towards cloud-centric distributed database evaluation

Abstract

The evolvement of cloud computing pushed the rethinking of traditional web service architecture from monolithic structures to distributed systems, which will benefit from the cloud offers such as resource pooling or rapid elasticity. Whereas the distribution of the mostly stateless business logic services fits well for distribution in the cloud, the distribution of stateful database services is more challenging. Hence in parallel to the evolvement of cloud computing, distributed databases moved in the focus of academia and industry with the result of variety of distributed database systems, which can be classified in relational databases, NoSQL and NewSQL database systems.

Theoretically the current representatives of these three database system classes claim to provide elasticity and “unlimited” horizontal scalability. As the characteristics elasticity and scalability match the cloud offerings, distributed databases seem to be a perfect match for implementing Database-as-a-Service systems (DBaaS). However, academia and industry have already proven significant differences in the elasticity and scalability of distributed databases in specific scenarios. As the cloud stack adds adds another level of complexity to the evaluation of distributed databases, an advanced evaluation framework is required to enable comparable and reproducible evaluations of distributed databases in the cloud.

In this context a new approach towards a cloudcentric evaluation framework for distributed databases is proposed, encompassing a model-driven evaluation methodology, an evaluation execution framework and a cloud-centric knowledge base for distributed database scalability and elasticity.