A performance comparison between parallel and lazy fetching in vertically distributed cloud databases

Abstract

The following work focuses on the severe performance issues that come along with the vertical data distribution approach of the SeDiCo framework. SeDiCo (A Secure and Distributed Cloud Data Store) is a framework that vertically distributes database data across several clouds in order to maintain data security, protection and privacy. However, this vertical data distribution requires joining data together before any client access. This join is extremely time-consuming and thus, this approach is currently not feasible in practical usage scenarios. In order to overcome those performance issues, this work proposes the introduction of a data cache architecture and two corresponding fetch procedures, i.e. lazy and parallel fetching. We further implemented both strategies, measured their performance and compared them to each other. Our results show that the parallel fetch strategy by far outperforms the lazy fetch strategy. Above that, this work outlines the entire caching architecture of our approach and discusses the two strategies in detail. Finally, this work concludes the measured results and discusses further challenges in the context of the SeDiCo framework as future work tasks.