Maintenance of the logical consistency in Cassandra

• Autores: Pablo Suárez-Otero González
• Directores de la Tesis: Pablo Javier Tuya González (dir. tes.)
• Lectura: En la Universidad de Oviedo ( España ) en 2023
• Idioma: español
• Tribunal Calificador de la Tesis: Antonio Vallecillo Moreno (presid.) , Raquel Blanco Aguirre (voc.) , Jesús Joaquín García Molina (voc.) , María José Suárez Cabal (voc.) , Paolo Atzeni (voc.)
• Programa de doctorado: Programa de Doctorado en Informática por la Universidad de Oviedo
• Menciones:
  • Internacional
• Calificación: Sobresaliente Cum Laude
• Enlaces
  • Tesis en acceso abierto en: RUO
• Resumen

  • Traditionally, software projects have exclusively used relational DBs (databases). However, in recent times the amount of data that these projects manage have exponentially increased, negatively affecting the DB performance. Due to this, different technologies specialized to manage big data have emerged, being one of these technologies the NoSQL DBs. One type of these DBs are the column-oriented DBs, whose schema contains tables without relationships between each other. Each one of these tables is designed specifically for satisfying the requirements of a query that will be executed by the client application, using a conceptual model during this design. This design strategy implies a denormalized schema that encourages data duplication, as the same data can be queried several times, thus storing the data in each of the tables that satisfy the queries. This brings problems regarding the logical consistency of the data, which ensures that the same data that are stored in two different tables are actually the same data. In this thesis we address two problems from column-oriented databases: 1) the maintenance of the logical consistency or data integrity when there are changes in the data and 2) the maintenance of the consistency between conceptual model and schema when there is an evolution of the DB. The first problem will be addressed from two different approaches: a preventive one named MDICA that prevents the creation of inconsistencies when modifying data in the DB and a reactive one named CONCODA that detects if inconsistencies have been created after modifying data in the DB. The second problem will be addressed through a MDE (model-driven engineering) approach named CoDEvo that will determine how to evolve the schema after the conceptual model changes in order to maintain the consistency between it and the conceptual model, which we name as inter-model consistency. MDICA determines the database statements required to perform against the database a given modification of data, which consists on a tuple to insert, delete, or update at the conceptual level, while maintaining the logical consistency. MDICA helps developers in two ways: generating the statements needed to maintain data integrity and producing messages to avoid problems such as loss of information, redundant repeated data, or gaps of information in tables. We have performed experimentation to validate MDICA, exhaustively checking the database operations that MDICA determined for a set of changes at the conceptual level for evaluation. CONCODA determines if a column-oriented DB maintains the logical consistency after one or more modifications of data have been performed against the DB. We have devised an oracle using a relational DB that implements the conceptual model, due to the fact that relational databases ensure logical consistency through the normalization of the schema and the implementation of integrity constraints. After the modifications of data are performed against both the relational DB and the column-oriented database, equivalent queries will be executed against both DBs. If these queries return the same data, then the logical consistency is ensured. CONCODA also serves as validation for MDICA to ensure that the database statements determined by MDICA actually maintain the logical consistency. The objective of CoDEvo is to obtain the changes that need to be applied to the schema when the conceptual model evolves in order to maintain consistency between the conceptual model and the schema. We have analysed several open-source projects to obtain patterns of how the schema evolves to automate this process. This part of the thesis will be integrated into a framework that will address other problems related to database evolution, including the maintenance of logical consistency.