MIGRATION FROM RELATIONAL DATABASE TO COLUMN-ORIENTED NOSQL DATABASE

• Main Author: Anggita S. Siregar, Rizky
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74534
• Institution: Institut Teknologi Bandung

Relational database conditions that start to get overwhelmed when processing large volumes of data and provide deteriorating performance, can be overcome by using a column-oriented NoSQL database. This is due to the advantages possessed by column-oriented NoSQL databases, such as scalability, performance, etc. Column-oriented NoSQL databases also have the advantage of being able to run analytical queries quickly. Because applications that use relational databases are already running, it is not possible to use a new NoSQL database without existing data in the relational database. Because of these needs, there is a need to migrate from relational databases to column-oriented NoSQL databases. In this final project, two alternative solutions are developed, namely direct and indirect methods, both of which consist of two stages, namely schema migration and data migration. The direct method performs a one-to-one mapping process, namely a table in a relational database will be migrated into one column family. The indirect method mainly focuses on two sub-processes, namely ERD extraction and ERD conversion to logical schema NoSQL columns. The ERD extraction process was developed from the research of Chiang et al. (1994) which utilizes metadata and inclusion dependencies. The ERD conversion process to a logical schema uses the Poffo & Mello algorithm (2016). The data migration stage consists of data extraction, data structure transformation, and data upload. The migration process using the indirect method is then implemented in an application, which was developed based on the application from Al-Ghifari & Azizah (2022). Based on the test results, the resulting two migration methods can migrate data from a relational database (MySQL) to a columnar NoSQL database (Cassandra) while maintaining data integrity. The test results also show that the indirect method has better performance than the direct method in all aspects, except for the data update case. It was also shown that applications developed for migration using the indirect method can be used to migrate MySQL databases to Cassandra.