CONICET | Buscador de Institutos y Recursos Humanos

Cloud computing (CC) environments can be analyzed at two different levels: infrastructure and application. At application level, the software architect defines the set of web applications that will be available in the cloud. Several authors have used simulation techniques on software architectures in order to predict and estimate quality properties of a software product [1]?[5]. Recently similar strategies have been used in CC at infrastructure level [6], [7]. However, all these approaches try to compute one (in some cases, two) quality aspects (leaving aside the rest of the properties).Quality evaluation by simulation at application level involves two different (but related) issues. First: definition of the quality aspects to be considered around the product. That is, the specification of which properties must be measured in order to get the quality level required. Second: deployment of a Simulation Model (SM) that provides a way to analyze the previous quality specification. This SM should not be built by the software architect given that it could confuse its vision about the product. Then, an analysis model that integrates both aspects is required.To solve this problem, the proposal is use an ontological approach to support the quality definition and a set of discrete event simulation models based in standard architectural components. A Quality Scheme (QS) supported by ISO/IEC 25010 [8] has been already defined [9], giving a mechanism that helps to understand quality by using software measurement strategies. The information recovered in a QS can be analyzed to get the lower level of data required. Ideally, these data should be obtained from the SM outputs. By developing an architectural design based on standard components; the SM can be built automatically. The proposed set of architectural components has been already defined in the metamodel presented in [10] (using architectural patterns detailed in [11]). The tool that support this metamodel allows develop valid software architectures. Mapping each component with an specific DEVS model [12]?[14] is the next step in this direction.