Analysis and evaluation of the riak cluster environment in distributed databases

• Published in: Computer standards and interfaces 2020-10, Vol.72, p.103452-11, Article 103452
• Main Authors: Topcu, Ahmet Ercan, Rmis, Aimen Mukhtar
• Format: Article
• Language: English
• Subjects: Basho-bench; Big data riak; Cluster Structured data; Computer networks; Experiments; Nosql database; Unstructured data
• ISSN: 0920-5489; 1872-7018
• DOI: 10.1016/j.csi.2020.103452

•Riak key-value database using Basho-bench for benchmarking performance.•Riak validated against NoSQL database in a distributed environment.•Riak compared performance with different sizes & scenarios of workloads.•Riak performance measured in terms of throughput and latency.•Riak results shows better correlation of performance and scalability. Many institutions and companies undergoing technological developments have been producing large amounts of structured and unstructured data. Special databases are required to deal with such data and NoSQL databases have thus emerged. They are widely used in cloud databases and distributed systems. In the era of big data, these databases provide a scalable solution with high availability. In this context, we need new architectures in order to store more and more different kinds of data. To obtain a good structure for large and diverse data, structures must be tested and analyzed in depth with the use of different benchmark tools. In this paper, we test the Riak key-value database to measure its performance in terms of throughput and latency, where huge amounts of data are stored and retrieved in different sizes in a distributed database environment. The throughput and latency of the NoSQL database in different types of experiments and with different sizes of data are compared. As we increase the data size in the experiments, an increase in the number of threads leads to better throughput and latency factor was reduced. High performance results are obtained for only read operations for all experiments. We observed that performance advanced when there was only read operations compared with a mix of read and update operations. Moreover, our findings intensify the understanding of the distributed database and have to help future developers through the experimental results shown in this paper.