Artisan Software has actively participated and at times led several activities related to the promulgation and standardization of modeling techniques and grammars for the specification and implementation of embedded, time and resource constrained systems. In this brief presentation, we will review Artisan Software's historical participation and report on the company's commitment to the MARTE Profile.
 

MARTE has been voted at OMG since June 2007 and offer the capabilities to design real-time and embedded systems with UML 2.1. By extending MagicDraw with this new standard, NoMagic is preparing a dedicated tool for Real-Time and embedded systems designers.

This talk is an overview of Edgewater’s RTEDge™ toolset, a modeling platform for proof based engineering of safety/mission critical hard real-time distributed systems. Specifying and building safety/mission critical real-time systems requires proving with high assurance the correctness of some functional and non-functional system requirements subsets deemed to have a major safety/mission impact. Taking a similar approach, RTEdge defines a small AADL and UML 2.0 modeling constructs subset which permits proof by static analysis of timeliness properties in hard real-time applications and enables the use of formal methods model checking techniques. The RTEdge modeling subset maps naturally into the MARTE extensions to UML 2.0
 

MARTE is a successor on the UML profile for Scheduling, Performance and Time (SPT), TriPacific has already developed a solution around SPT, then it is a natural way to now address a solution around MARTE.

MARTE is important, however if it is not used in conjunction with "some" analysis tool or capability, then it is only documentation. This is not bad, however it is not as powerful as with the use of Performance analysis or Schedulability analysis. The “MARTE – RapidRMA” bridge provides the type of solution needed to make MARTE useful beyond documentation.
 

Component-based software is in now a key technology to develop systems in the embedded market, e.g. telecom, Software Defined Radio, automotive, and aerospace. One of the key advantages of the component-based approach is that it allows decoupling applications, which greatly increases reuse and portability. In this presentation, we will discuss how the Zeligsoft Component-Driven Software Engineering solution combined with MARTE can be used to specify non-functional properties and to analyze key aspects of complex embedded systems such as deployment, resource usage, and timing.
 

USER PRESENTATION/CASE STUDIES

To build applications that achieve their performance goals, performance must be considered as early as possible in application design. Hence, design models at any level of abstraction should be analyzed with regard to the performance of the system modeled. Yet, designers do not typically have the performance expertise required for such analysis. In this talk, we present model-driven performance engineering, an approach developed in the MODELPLEX project that seamlessly couples performance simulation with MARTE-annotated UML models to allow push-button performance analysis without requiring deep performance expertise of designers.
 

MARTE has been experimented in the automotive domain inside the European project ATESST. This presentation will show you the main results of this experimentation and the link between MARTE and the automotive langage EAST-ADL.
 

Processors now have more than one core, therefore it is essential to be able to design and master the complexity of the application and the multi-core processors. THALES has experimented with the capabilities of MARTE to design a signal processing application and hardware with a multi-core processor (Processor CELL). This presentation will show you that with MARTE, it is possible to use a classical UML tool with MARTE to achieve this objective.
 

ADVANCED TECHNOLOGIES

One key challenge of MARTE is to offer the tool support for making user models productive within a consistent development process. This presentation shows our current results to integrate a set of tools enabling practitioners for incremental timing analysis, code generation, and prototype execution. In this approach, MARTE-annotated models, supporting a particular model of computation, are executed and simulated in a real-time framework, and iteratively tuned (e.g., task allocation) by schedulability analysis tool results.
 

MARTE will be important for designing software with soft real-time specifications, such as average delay or 95th percentile delays, as well as for embedded systems. Network services can be extremely time-sensitive, for example for financial trading, multimedia delivery, or distributed games. Performance analysis is used for these cases, and Layered Queueing is a performance model designed for analyzing complex, large-scale layered systems of servers and networks. Layered queueing models are derived automatically by transformation from annotated UML specifications, and can be solved by the LQNS solver. The same infrastructure supports transformation to other performance models such as queueing networks or timed petri nets.
 

The Schedulability Analysis sub-profile in MARTE provides a number of constructs for the creation of scenario based models suitable for schedulability analysis with modern techniques. This presentation shows the mapping strategy and current advances in the generation of the models necessary to perform the analysis with the Modeling and Analysis Suite for Real-Time Applications (MAST), an open source effort made in the University of Cantabria to disseminate the least pessimistic offset-based analysis techniques, and enhance the state of the practice at industrial level.