A method for modeling a schema for graph databases

In (Roy-Hubara, & et al, 2017), we introduced a method for modeling a graph database schema, using a rich ERD that represents the domain of an application. The ERD is mapped in two steps to a graph database schema (GDBS) that preserves the semantics and constraints defined in the original conceptual schema. In the order to evaluate the proposed method we conducted a controlled experiment in which we compared it with alternative methods, measuring differences in quality of the GDBS created by designers, time to complete the design task, and designers' satisfaction with the modeling method. The results of the experiment, which are reported in Roy-Hubara, & et al, 2018), showed differences in the quality of the created GDBS: subjects who used the proposed modeling method produced correct schemas, while subjects who used the ERD only made some errors, but fewer errors than the subjects of the control group who used "ad-hoc" method. With respect to time, we found significant differences: subjects who used the proposed method spent significantly more time than the subjects who used the other two methods. In the future, we plan to repeat the controlled experiment, using more and homogeneous subjects, who are more trained with graph databases, and utilize design tasks of different size and complexity. We also plan to implement the proposed method, i.e., to create a domain independent tool to be used as plug-in for graph database systems, that will enable to define a (initial) schema of the application. Such implementation might be adopted by graph database system providers. Finally, we intend to extend the proposed schema modeling method to model other types of the NoSQL databases.