Data Modeling Guidelines for NoSQL Document-Store Databases

Abstract

Good database design is key to high data availability and consistency in traditional databases, and numerous techniques exist to abet designers in modeling schemas appropriately. These schemas are strictly enforced by traditional database engines. However, with the emergence of schema-free databases (NoSQL) coupled with voluminous and highly diversified datasets (big data), such aid becomes even more important as schemas in NoSQL are enforced by application developers, which requires a high level of competence. Precisely, existing modeling techniques and guides used in traditional databases are insufficient for bigdata storage settings. As a synthesis, new modeling guidelines for NoSQL document-store databases are posed. These guidelines cut across both logical and physical stages of database designs. Each is developed based on solid empirical insights, yet they are prepared to be intuitive to developers and practitioners. To realize this goal, we employ an exploratory approach to the investigation of techniques, empirical methods and expert consultations. We analyze how industry experts prioritize requirements and analyze the relationships between datasets on the one hand and error prospects and awareness on the other hand. Few proprietary guidelines were extracted from a heuristic evaluation of 5 NoSQL databases. In this regard, the proposed guidelines have great potential to function as an imperative instrument of knowledge transfer from academia to NoSQL database modeling practices.

DSP: Schema Design for Non-Relational Applications

“…The model proposed in this paper adopts the guidelines as proposed by [70]. These guidelines (in Table 3) are helpful in prioritizing relationships.…”

Automatic schema suggestion model for NoSQL document-stores databases

“…Firstly, the IS was generated through the proposed model which adopted the new cardinalities and styles [11]. The proposed model also incorporated modeling guidelines proposed in [12]. Secondly, the proposed mathematical formulas do not focus on calculating only cluster or document availability but also provide consistencies across schemas.…”

“…Semantic mappings of the recorded entities are then initiated by the model itself starting from third layer. This is achieved partly using modeling guidelines [12] and new generation cardinalities [11]. The following algorithms explain the logic.…”

“…The said earlier studies were motivated by the industrially proven problems, theoretically grounded with basic modeling facts, and empirically validated through repeated and rigorous experiments. One of the aforesaid studies, focuses on cardinality notations and styles [11], while the other presented modeling guidelines [12]. The two studies scoped at NoSQL document-store databases.…”

“…With the above described model, parameters together with their associated data can be supplied into the proposed model to produce initial schema based on solid theoretical and empirical foundations adopted from [11,12]. It should be unequivocally noted that, the proposed model does not produce fully functional schema which can be 100% copy-pasted, rather produce a schema with basic modeling concepts that can be used at the initial stage of database design.…”