1. Henry Muccini - Computer Science Department, Universita' dell'Aquila, Italy (muccini@di.univaq.it) Paola Inverardi - Computer Science Department, Universita' dell'Aquila, Italy (inverard@di.univaq.it) Patrizio Pelliccione - Software Engineering Competence Center, University of Luxembourg(patrizio.pelliccione@uni.lu) DUALLY: Putting in Synergy UML2.0 and ADLs SEAGroup

2. 2 Modeling Software Architecture  Two main classes of languages have been used so far to model software architectures:  formal architecture description languages (ADLs)  model-based specifications with UML

3. 3 Still supported ADLs Very limited industry buy-in to date

4. 4 Problems with existing ADLs 1/2  ADLs tend to focus on a single area of interest  analysis (Wright)  refinement (SADL)  dynamism (Weaves)  Within these areas ADLs tend to direct their attention to a particular technique  Wright analysis for deadlocks  They leave other facets unexplored

5. 5 A possible scenario C1 C2 C5 C4 C3 ADL1 Deadlock analysis V Performance analysis ADL2 X C1 C2 C5 C4 C3 C2’’C2’ The deadlock model needs to be manually adjusted and re-analyzed

6. 6 Problems with existing ADLs 2/2  Specialized semantic basis  High degree of formality  Limited tool support  Lack of lifecycle- wide support  Lack of tie to requirements  Limited mechanisms to preserve architectural properties in detailed designs and implementations  Limited support for architecture-based evolution

7. 7 UML for SA modeling  Industries still tend to prefer model- based (semi-formal) notations  UML is emerging as the de facto standard design notation of choice in industrial software development  Many extensions and profiles have been proposed to “adapt” UML to model architectures

8. 8 UML for SA modeling: problems  These extensions permit to reduce the gap between UML and ADLs, but they still fail in representing all aspects of ADLs  modeling for documenting is quite different from modeling for analysis, and different analysis techniques usually require different notations  Any time a new slightly different analysis is required, new modeling concepts are needed  There is neither a unique language for representing SAs, not a unique fit between UML and ADLs

9. 9 Can we really advocate a stronger synergy between UML and ADLs can be created?

10. 10 Dually: Putting in Synergy UML 2.0 and ADLs Identify a core set of architectural elements always required Create an UML profile able to model the core architectural elements previously identified Provide extensibility mechanisms to add modeling concepts needed for specific analysis Describe how semantic links mechanisms can be kept between different notations

11. 11 Dually Profile ConceptStereotypeBase Class Architectural Component >Component Relations among SA components >Depending Connectors >Component Channels >Assembly Connector Package for State Machines & Interaction Diagrams >Package

12. 12 Extensibility Mechanisms: weawing meta-models M Dually M FT Overrides OperatorInherits OperatorRename Operator

13. 13 Extensibility Mechanisms: woven model

14. 14 ROSATEA 2006 The Role of Software Architecture Testing and Analysis http://www.di.univaq.it/Rosatea2006/ Engineering Fault-Tolerant Systems - EFTS 2006