Conversion of Entity-Relationship Model to NoSQL Document-Oriented Database Logical Model Using Workload Information and Entity Update Frequency

Although dubbed schema less, NoSQL databases, especially the document-oriented ones, can still make use of effective data modeling techniques in order to provide high-performance databases. In this paper, we present a proposal of converting Entity-Relationship Model (ER Model) into NoSQL document-oriented database logical model based on a workload-driven logical modeling algorithm by Lima & Mello [6]. This work improves the algorithm by considering, not only database workload, but also entity update frequency. The workload information and the entity update frequency are used to determine whether to refer or to embed in the logical model. Embedding or referencing are two prominent “relationships” among collections in document-oriented databases. We tested the improved conversion algorithm into a case study and the performance of the resulting logical model was tested on MongoDB in comparison to the logical model resulting from the original algorithm. The performance tests show that the logical model resulting from the improved algorithm provides better performance in read operations, but worse performance in create and update operations in collections with embedded relationships.