Multi-model Database Management Systems - A Look Forward

Abstract

The existence of the variety of data models and their associated data processing technologies make data management extremely complex. In this paper, we envision a single Multi-Model DataBase Management Systems (MMDBMS) providing declarative accesses to a variety of data models. We briefly review the history of the evolution of the DBMS technology to derive requirements of MMDBMSs and then we illustrate our ideas of building MMDBMSs satisfying those requirements. Since the relational algebra is not powerful enough to provide a mathematical foundation for MMDBMSs, we promote the category theory as a new theoretical foundation, which is a generalization of the set theory. We also suggest a set of shared data infrastructure services among data models to support "Just-In-Time" multi-model data access autonomously.

A Formal Category Theoretical Framework for Multi-model Data Transformations

“…Leclercq et al [18] built foundations of polystores on the tensor-based data model. Liu et al [19] visioned that the foundations of multimodel databases could be built on category theory and we continue this work for polystores and multi-model databases.…”

Unified Management of Multi-model Data

“…Regardless of the adopted strategy, the theory of categories [14], associative arrays [11], or description logics [1] could be utilized to internally model the data in a formal, abstract, and rigorous way. Complex non-relational systems often involve a variety of heterogeneous and interrelated models -models that are, unfortunately, expressed using several modeling languages.…”

“…Considering the environment of multi-model data, one application can store data in one data model, whereas later the same data can be queried by another application using a different model via multi-model data views [14]. Hence, multi-model data transformation can exploit the genuine value of multi-model databases which enable applications requiring different data models to share the same platform.…”

Empowering big data analytics with polystore and strongly typed functional queries

“…ETL processes are expensive and time consuming tasks, and transforming multiple datasets into a single data model in a data warehouse can impact negatively performances and reduce the expressivity of the original data model. In-database analysis cannot take easily into account new algorithms, as they require a specific development for each database model in order to fit the data structure required by the algorithm [26,27]. So researches on polystores are directed towards ETL streaming systems [13,41], multi-database query language [14], unification models [15], and parallel query processing as well as the integration of polystore with analysis frameworks [22].…”

“…Nevertheless, data processing pipelines are becoming complex and heterogeneity is no longer observed only at the data level but also at the analysis level [2]. Developing a pivot data model which can generalized polystore underlying data models and which can facilitate data transformations to feed quickly algorithms implemented with frameworks is a major challenge [17,19,26].…”