Comparison of Database and Workload Types Performance in Cloud Environments

The rapid growth of unstructured data over the last few years, has led to the emergence of new database management systems. Traditional relational databases, despite their wide adoption and plethora of features, begin to show weaknesses when having to deal with very large amounts of data. Numerous types of databases have emerged in the Cloud domain, in order to exploit the elasticity of Cloud environments, while relaxing the typical ACID considerations and investigating trade-offs of the CAP theorem. The aim of this paper is to investigate how such offerings (MongoDB, Cassandra and HBase namely), based on these tradeoffs, behave when deployed in virtual environments (of the BONFIRE facility) and how they are measured against widely used benchmarks such as YCSB. The results may be helpful for potential adopters to choose from these offerings, based on their individual needs for specific workloads or query structures.