Multi-model Database Design and Query Processing in NewSQL DBMSs

What is it about?

NewSQL DBMSs are hybrid systems. Their advent can help to promote new types of DBMSs with the required capability to better address the requirements of modern applications compared to SQL DBMSs and NoSQL DBMSs. To achieve this, they must focus on two key objectives: (1) Maintain the advantages of SQL DBMSs by offering: (i) capabilities for managing fixed-schema collections of values while ensuring data integrity and independence; (ii) a non-procedural interface language that facilitates query optimization; (iii) a transaction execution environment that guarantees the Atomicity, Consistency, Isolation, and Durability (ACID) properties. (2) Consider the advantages of NoSQL DBMSs, which include: (i) flexibility through flexible-schema or schema-less collections of values, adhering to various structures (e.g. structured, hierarchical-and-structured, hierarchical-and-semi-structured, unstructured, or graph-oriented); (ii) customizable data logical and physical organization enabling optimal query latency; (iii) unlimited horizontal scaling to accommodate workload changes; (iv) continuous data availability. To our knowledge, there is currently no research on comprehensive methods for designing a database specifically aimed at DBMSs combining all these advantages, with the following objectives: (1) identifying and defining heterogeneous collections of values, adhering to different data models, for storage and manipulation within only one database, ensuring they meet the needs of a business's operations; (2) guaranteeing latency, scaling capability, and continuous data availability. This paper outlines such a coherent and comprehensive methodological approach for designing a database specifically intended for a NewSQL DBMS.

Featured Image

Photo by Favour Usifo on Unsplash

Photo by Favour Usifo on Unsplash

Why is it important?

Many of today's NewSQL DBMSs are currently under development by leveraging a proven SQL DBMS engine for incorporating well-known concepts and functionalities under the relational model to provide the NoSQL DBMSs engines. Following this approach, these DBMSs introduce a broad range of possibilities, thereby increasing the complexity of database design. Furthermore, some of these possibilities open the door to very poor practices. Therefore, to define such a design approach, we rely primarily on the ANSI/SPARC architecture of schemas. The resulting methodology aims to help the designer fully describe a NewSQL database across multiple levels of abstraction, using complementary yet consistent or orthogonal schemas, supported by schema-mapping techniques. Each level of abstraction enables the designer to focus on the objectives specific to that level by relying on the most appropriate concepts, formalisms, techniques, and functionalities required.

Perspectives