Software Product Line (SPL) engineering is a software development approach that takes advantage of the commonality and variability between products from a family, and supports the generation of specific products by reusing a set of core family assets. This paper proposes a UML model transformation approach for software product lines to derive a performance model for a specific product. The input to the proposed technique, the "source model", is a UML model of a SPL with performance annotations, which uses two separate profiles: a "product line" profile from literature for specifying the commonality and variability between products, and the MARTE profile recently standardized by OMG for performance annotations. The source model is generic and therefore its performance annotations must be parameterized. The proposed derivation of a performance model for a concrete product requires two steps: a) the transformation of a SPL model to a UML model with performance annotations for a given product, and b) the transformation of the outcome of the first step into a performance model. This paper focuses on the first step, whereas the second step will use the PUMA transformation approach of annotated UML models to performance models, developed in previous work. The output of the first step, named "target model", is a UML model with MARTE annotations, where the variability expressed in the SPL model has been analyzed and bound to a specific product, and the generic performance annotations have been bound to concrete values for the product. The proposed technique is illustrated with an e-commerce case study. Copyright 2008 ACM.

MARTE., Model transformation, Software Performance Engineering, Software Product Line, UML
7th International Workshop on Software and Performance 2008, WOSP'08
Department of Systems and Computer Engineering

Tawhid, R. (Rasha), & Petriu, D. (2008). Towards automatic derivation of a product performance model from a UML software product line model. Presented at the 7th International Workshop on Software and Performance 2008, WOSP'08. doi:10.1145/1383559.1383571