Master of Science (MS)
We witnessed a dramatic increase in the volume, variety and velocity of data leading to the era of big data. The structure of data has become highly flexible leading to the development of many storage systems that are different from the traditional structured relational databases where data is stored in “tables,” with columns representing the lowest granularity of data. Although relational databases are still predominant in the industry, there has been a major drift towards alternative database systems that support unstructured data with better scalability leading to the popularity of “Not Only SQL.”
Migration from relational databases to NoSQL databases has become a significant area of interest when it involves enormous volumes of data with a large number of concurrent users. Many migration methodologies have been proposed each focusing a specific NoSQL family. This paper proposes a heuristics based graph transformation method to migrate a relational database to MongoDB called Graph Transformation with Selective Denormalization and compares the migration with a table level denormalization method. Although this paper focuses on MongoDB, the heuristics algorithm is generalized enough to be applied to other NoSQL families. Experimental evaluation with TPC-H shows that Graph Transformation with Selective Denormalization migration method has lower query execution times with lesser hardware footprint like lower space requirement, disk I/O, CPU utilization compared to that of table level denormalization.
Mullapudi, Krishna Chaitanya, "Schema Migration from Relational Databases to NoSQL Databases with Graph Transformation and Selective Denormalization" (2019). Master's Projects. 714.