Performance comparison of relational and NoSQL database systems in the cloud using an example from the field of e-commerce

E-commerce is a growing market. Companies running e-shops are faced with the task of choosing a database. Cloud database platforms are not much explored in this aspect. The most used databases are relational and NoSQL, which is why this work focused on them. This paper's primary goal is to perform a performance comparison of selected relational and NoSQL database management provided in the form of a Database as a Service, using an example from the field of e-commerce. The thesis analyzes the current mainstream offer of cloud database services. For performance testing, the work selects CosmosDB for PostgreSQL (relational) and CosmosDB for MongoDB (NoSQL) running on the Microsoft Azure platform. The paper uses a subset of anonymized data from a real system, using the example of a data model from the field of e-commerce. It uploads it to both databases using a custom Java application. The thesis compares the tested systems using the load testing method with the JMeter library. The measurements show that the average time of all operations for MongoDB was always many times higher than the time of PostgreSQL. So, it can be concluded that PostgreSQL is more powerful. However, the exception is the read operation, where MongoDB is faster. MongoDB also achieves a lower error rate for a more extensive range of queries.