Modeling Apache Hive based applications in Big Data architectures

Performance prediction for Big Data applications is a power- ful tool supporting designers and administrators in achieving a better exploitation of their computing resources. Big Data architectures are complex, continuously evolving and adap- tive, thus a rapid design and verification modeling approach can be fit to the needs. As a result, a minimal semantic gap between models and applications would enable a wider number of designers to directly benefit from the results. The paper presents a multiformalism modeling approach based on a one-to-one mapping of Apache Hive querying primitives to modeling primitives. This approach exploits a combina- tion of proper Big Data specific submodels and Petri nets to enable modeling of conventional application logic.