Prof. Ernesto Damiani gave a keynote at the 11th Int. Conf. on Evaluation of Novel Approaches to Software Engineering (ENASE2016) titled “TOWARDS MODEL-DRIVEN BIG-DATA-AS-A-SERVICE”.

Much work has been done to show that model-driven development—following the classic Model-Driven Architecture (MDA) approach—is also advantageous in the context of data intensive applications, which have become the norm in key business domains like CRM, sales/marketing, HRM and Business Process analysis. As a result, a number of methodologies have been proposed for developing data-intensive applications via a chain of model transformations. Recently, however, the advent of Big Data has brought on a new fundamental issue. While model transformations usually introduce each architecture dependent feature at a fixed point of the model refinement chain, Big Data computations can only fulfil their promises of scalability and efficiency if binding architectural and data modelling decisions are taken at the last possible moment, i.e. when information on the data distribution, volume and nature becomes available.
In this talk, it as been argued that Big-Data-as-a-Service, where data features can be different for each deployment of the data pipeline, can benefit from a Software-Product-Line (SPL) parametric approach to keep multiple alternative models alive and postpone binding modelling decision in order to make at at the “right” (i.e., the last possible) moment. The traditional model transformations has been compared with a “lazy” ones and discussed how to delay and parametrize modelling decisions for all key aspects of a Big Data application, including data modelling, parallelization and visualization.