Performance Model Building of Pervasive Computing∗

Performance model building is essential to predict the ability of an application to satisfy given levels of perfor-mance or to support the search for viable alternatives. Using automated methods of model building is becom-ing of increasing interest to software developers who have neither the skills nor the time to do it manually. This is particularly relevant in pervasive computing, where the large number of software and hardware compo-nents requires models of so large a size that using tradi-tional manual methods of model building would be error prone and time consuming. This paper deals with an automated method to build per-formance models of pervasive computing applications, which require the integration of multiple technolo-gies, including software layers, hardware platforms and wired/wireless networks. The considered performance mod-els are of extended queueing network (EQN) type. The method is based on a procedure that receives as in-put the UML model of the application to yield as output the complete EQN model, which can then be evalu-ated by use of any evaluation tool. 1