Modelagem de desempenho do banco de dados cassandra

Abstract

Performing large amount of data storage is a challenge since the demand to manage the massive volume of data is growing. Among the reasons that provide this growing, there are social networks, Internet of things and cloud computing. In this context, new data storage systems have been emerging, for instance, non-relational databases NoSQL. These databases are been used by companies that manage massive volume of data attending their demands on scalability and high availability. One of the main strategies for choosing a database is to evaluate the performance of typical operations like insert and read data through measurement techniques. However, this technique demands time and investment to choose the suitable database for a specifi c application. An alternative to evaluate the performance of databases, saving investment and time, is through modeling techniques. Considering such a scenario, this work evaluates the performance of a NoSQL database using simulation and modeling techniques. In this case, the response time in insertion operations is evaluated. The NoSQL database evaluated is Cassandra and the performance model was developed in stochastic Petri net. In order to validate the performance model, measurements were conducted on the environment with the support of the Yahoo! Cloud Serving Benchmarking tool. The results demonstrate that the proposed performance model can be adopted to analyze the performance of Cassandra in data insertion process.