Multi-Model database systems combine the advantages of traditional and NoSQL database systems. However, the management of these systems is challenging, as users have to design an appropriate storage strategy for their data. One of the most influential factors in the storage strategy is the selection of indexes. Indexes can significantly improve query performance, but they require additional storage space and maintenance overhead. Index selection problem is well-studied in the context of single-model Database Management Systems (DBMSs), but there is a lack of research in the context of multi-model database systems. We address this problem by conducting a survey of current state-of-the-art index selection algorithms and evaluating their applicability to other DBMSs. The results reveal the strengths and weaknesses of existing algorithms and highlight the need for specialised algorithms for multi-model database systems. Moreover, we formulate open questions and suggest future research directions in this field. Our research provides a foundation for the development of efficient index selection algorithms for multi-model DBMSs.

Multi-model data encompasses structurally distinct data, including relational, document, graph, key/value, columnar, etc., managed within a single system, such as a multi-model database or a data lake. Querying multi-model data requires strategies that balance unification and integration across diverse models and query languages. This paper presents DortDB, an extensible framework enabling cross-model queries combining well-known query languages and offering intuitive flexibility and optimization via a unified algebra. Though a small-scale in-memory prototype is to be demonstrated, its principles can be extended to distributed systems.

Functional dependency (FD) discovery is fundamental in data profiling. Inevitably, existing approaches can return fake FDs that hold only coincidentally. Moreover, these approaches fall short of identifying ghost FDs that would be observable in a clean dataset, but that remain undetected because of outliers in the data. We introduce an interactive method for dependency discovery that augments an Armstrong relation with additional tuples. We rely on artificially generated negative examples that emulate real-world tuples to help expose fake FDs. In addition, we rely on domain experts to confirm that positive examples indeed reflect the characteristics of the original dataset. Our tool prototype FDepHunter thus provides a novel human-in-the-loop workflow where the set of discovered FDs can be iteratively refined.

In this paper, we focus on the problem of evolution management of multi-model data. With the changing user requirements, the schema and the data need to be adapted to preserve the expected functionality of a multi-model application. We introduce a tool MM-evocat based on utilising the category theory. We show that the core of the tool, i.e., the categorical representation of multi-model data, enables us to grasp all the specifics of the individual models and their possible combinations. Its simple but powerful formal basis enables unique and robust support for evolution management.

Multi-model data is organised in various mutually interlinked formats and models, often with contradictory features. In addition, its structure may change over time, and its size can grow to the extremes of Big Data. In terms of research and practical processing, this creates one of the most complex challenges of effective data management.

In this vision paper, we introduce an idea of a framework that would enable us to model, represent, and manage multi-model data in a unified and abstract way. Its core idea exploits constructs provided by category theory, which is sufficiently general but still simple enough to cover any of the logical data models used in contemporary databases. Focusing on promising features and taking into account mature and verified principles, we overview the key parts of the framework and outline open questions and research directions that need to be further investigated. The ultimate objective is to pursue the idea of a self-tuning system that would permit us to collapse the traditionally understood conceptual and logical layers into just a single model allowing for unified handling of schemas, data instances, as well as queries.

Following the current trend, most of the well-known database systems, being relational, NoSQL, or NewSQL, denote themselves as multi-model. This industry-driven approach, however, lacks plenty of important features of the traditional DBMSs. The primary problem is a design of an optimal multi-model schema and its sufficiently general and efficient representation. In this paper, we provide an overview and discussion of the promising approaches that could potentially be capable of solving these issues, along with a summary of the remaining open problems.