NoSQL Scalability Performance Evaluation over Cassandra

Authors

Abstract

The implementation of Smart-Cities is growing all over the world. From big cities to small villages, information able to provide a better and efficient urban management is collected from multiple sources (sensors). Such information has to be stored, queried, analyzed and displayed, aiming to contribute to a better quality of life for citizens and also a more sustainable environment. In this context it is important to choose the right database engine for this scenario. NoSQL databases are now generally accepted by the database community to support application niches. They are known for their scalability, simplicity, and key-indexed data storage, thus, allowing an easy data distribution and balancing over several nodes.

In this paper a NoSQL engine is tested, Cassandra, which is one of the most scalable, amongst most NoSQL engines and therefore, a candidate for use in our application scenario. The paper focuses on horizontal scalability, which means that, by adding more nodes, it is possible to respond to more requests with the same or better performance, i.e., more nodes mean reduced execution time. Although, adding more computational resources, does not always result in better performance. This work assesses how each workload (e.g., data volume, simultaneous users) influence scalability performance. An overview of the Cassandra database engine is presented in the paper. Following, it will be tested and evaluated using the benchmark Yahoo Cloud Serving Benchmark (YCSB).

Keywords

NoSQL, Cassandra, Horizontal scalability, Performance, Database, YCSB, BigData, Distributed, Parallel, Smart-City, CityAction

Conference

International Conference Europe Middle East & North Africa Information Systems and Technologies to Support Learning, October 2018

DOI