Database Performance Comparison of Xen, KVM and OSv Using Cassandra

V16 technologies have promoted the fast development of large-scale information systems across all over the world. The distributed database access also has been improved to adapt the non-centralized virtual machine. In this paper, we utilize the typical distributed database Cassandra to evaluate the performance of different virtualization technologies. The evaluation selects KVM, Xen and OS^v as the target virtualization platforms to test the read/write latency in Cassandra database. The result shows that OS^v has the shortest read and write latency and reaches the best performance. Furthermore, we compare the architecture of the target virtualization platforms to analyze the relationship between applications and virtualization architectures. The analysis illustrate that more software-defined abstract levels in the virtualization architecture lead to worse system performance, because the level between the application and hardware becomes deeper, and then the program execution flow becomes longer. Finally, we demonstrate that the virtualization platform with thinner architecture levels has a better performance in the distributed data access, due to the specified full-virtuallzation, tidy hierarchical level, removal of duplicate protection and software-defined abstract.