Schema Design and Evaluation for Cassandra

Abstract

Cassandra is one of the widely used distributed database because of some featureslike fault tolerance, partition tolerance, seamless replication and scalability. However,designing the schema for Cassandra is a challenging task. This work focuseson automating the schema design procedure for Cassandra. Following are the importantreasons for schema design automation. 1) Schema design for Cassandrarequires query load as an input along with the conceptual model which is differentfrom relational schema design. 2) Inclusion of a query load as an input addsmore complexity to the schema design process. To automate the schema designprocess, we study and formulate some mapping rules and develop an algorithmcapturing these rules. The algorithm takes two inputs namely ER diagram andapplication queries. The algorithm produces an output schema which containsthe list of attributes required to answer the query, partition key attributes andclustering key attributes. Evaluation of the algorithm is done manually. We usethree case studies for this purpose, out of which two are from literature namelydigital library system and hotel management system and remaining one is ourown application which is an E-commerce application. Results which we get onexecuting this algorithm on mentioned cases are turned out to be correct.