Databases For Mediation Systems : Design and Data scaling approach

Abstract: Context: There is continuous growth in data generation due to wide usage of modern communication systems. Systems have to be designed which can handle the processing of these data volumes efficiently. Mediation systems are meant to serve this purpose. Databases form an integral part of the mediation systems. Suitability of the databases for such systems is the principle theme of this work. Objectives: The objective of this thesis is to identify the key requirements for databases that can be used as part of Mediation systems, gain a thorough understanding of various features, the data models commonly used in databases and to benchmark their performance. Methods: Previous work that has been carried out on various databases is studied as a part of literature review. Test bed is set up as a part of experiment and performance metrics such as throughput and total time taken were measured through a Java based client. Thorough analysis has been carried out by varying various parameters like data volumes, number of threads in the client etc. Results: Cassandra has a very good write performance for event and batch operations. Cassandra has a slightly better read performance when compared to MySQL Cluster but this differentiation withers out in case of fewer number of threads in the client. Conclusions: On evaluation of MySQL Cluster and Cassandra we conclude that they have several features that are suitable for mediation systems. On the other hand, Cassandra does not guarantee ACID transactions while MySQL Cluster has good support. There is need for further evaluation on new generation databases which are not mature enough as of now.