Performance Evaluation of NewSQL Databases in a Distributed Architecture

Abstract

In the last decade, application architectures have evolved drastically, moving from monolithic architectures to distributed architectures where deployment has shifted from dedicated on-premises servers to the cloud. Distributed architectures and cloud computing has enabled businesses to scale their application components across different geographical locations. While it is easy to scale the application layer, scaling its database layer that relies on traditional SQL databases is challenging and often is a common source of bottlenecks when it comes to application performance. This paper evaluates the performance characteristics between two NewSQL databases solutions, MySQL NDB Cluster vs. TIBCO ActiveSpaces IMDG. Serving as an application layer, a simulation of banking microservices, handling loyalty rewards, is developed to interface with these databases to derive additional insights and comparison. In the context of a geographically distributed architecture, our experiment results show that MySQL NDB Cluster is more suited for critical business transactions requiring ACID properties while TIBCO ActiveSpaces is better suited for use cases that require big data ingestion, such as IoT platforms.

Keywords

• Database performance
• NewSQL
• SQL
• MySQL NDB
• IMDG
• in-memory data grid