A Brazilian survey on UML and model-driven practices for embedded software development

Abstract

This paper brings statistical findings from a survey about the use of UML modeling and model-driven approaches for the design of embedded software in Brazil. The survey provides evidences regarding the maturity of use of UML and model-driven approaches, how they are employed, and which and where the professionals who use them are. Technical, organizational, and social aspects were investigated and documented by making use of a descriptive research method. Such aspects seemingly reflect the opinions of software engineers on how they perceive the impact of using UML and model-driven approaches on productivity and quality in embedded software development. Results show that most participants are clearly aware of the modeling approach value, even though they practice it only to a limited degree. Most respondents who make use of model-driven approaches attest that productivity and portability are the key advantages of their use.

Introduction

The term Model Driven Engineering (MDE) is used to describe model-driven software development approaches in which abstract models of the software systems are designed and systematically transformed into concrete implementations (France and Rumpe, 2007). Model Driven Architecture (MDA), a standard from the Object Management Group (OMG), is MDE's best known and currently adopted model-driven approach.

In MDA, the systematic use of models as engineering main artifacts occurs throughout the entire software development cycle (OMG, 2003). The primary technical advantages claimed by MDA consist of productivity, portability, reusability, and interoperability (García-Díaz et al., 2010, Kleppe et al., 2003). In this context, after the launch of the MDE approach, the role of the Unified Modeling Language (UML) became even more important, positioning the UML as the core part in the software development process. UML is a Software Engineering industry standard that stems from the principle of software development best practices (OMG, 2010).

Furthermore, several studies currently point out the importance of integrating model-driven approaches in the embedded software development process, aiming to minimize the effects of the heterogeneity of platforms and the increasing complexity inherent in this type of software (Jeon et al., 2009, Karsai et al., 2008). Embedded software comprises incorporated hardware features and is commonly subject to rigid constraints, which indeed represent a major challenge for the community of developers (Marwedel, 2006). The use of model-driven approaches for designing embedded software improves validation and verification as well as facilitating reuse and evolution (Espinoza et al., 2009).

This paper aims to identify gaps in the current comprehension of issues, such as knowledge of how exactly UML and MDE/MDA are used in industry; understanding of how social and organizational factors impact on UML and MDE/MDA use. More specifically, the following UML-related issues were investigated: UML diagrams more commonly used, UML complexity level under the view of the participants, and UML use maturity. Moreover, this paper emphasized the benefits of model-driven approaches in terms of productivity, portability, and quality with regard to the following activities: code generation, model transformations, documentation, and testing. In addition, MDE/MDA maturity of use in the organization was also investigated. Social and organizational aspects obtained were used in cross-tab reports so as to acquire new knowledge from the data collected. This paper compiles answers that illustrate aspects derived from the practical experience of software engineers who took part in a survey involving 209 Brazilian embedded software developers.

These aspects must be known because, despite the constant UML advancements, UML has been used in different ways and for different purposes in software development designs. Frequently, UML is used in the software development traditional context, not taking into account model-driven approaches. Therefore, models are used by programmers just as a reference for specifying and documenting software systems. In MDE/MDA approaches, however, models are used as the key artifacts in all software development stages. In this case, the development of software can be seen as a series of successive transformations from one model into another, including code generation from models.

The style and strictness in UML-based modeling also show a variation that relies on several aspects, such as the previous experience of the software engineers and the deadline defined for the software development. It is important to mention that UML is one of the most widely criticized modeling languages (France and Rumpe, 2007). As a reason for that, one could point out some of its deficiencies, for instance its lack of expressiveness (Pardillo and Cachero, 2010). Also, in many cases UML is not used, or its use is limited, on the grounds its complexity (Dori, 2002). In other cases, it is criticized due to the difficulty in modeling applications related to specific domains, e.g. embedded systems (Friedenthal et al., 2008).

The results obtained thus provide criteria for the identification of problems that hamper and impair the UML use for the development of embedded software in Brazil. The listed problems may be used by the academic community to suggest alternatives to the improvement of the UML usability, focusing on the future use of MDE/MDA approaches for embedded software design. Besides, the emphasis on embedded software is justified due to the wide range of existing applications such as industrial automation, electronic industry, telecommunications or military products. Currently, the growing complexity of embedded products is demanding the use of different software development approaches, such as the model-driven approach.

The paper is structured as follows. Section 2 brings the related works. Section 3 presents the details of the methodology used to perform the survey, including the purpose, instrument, and data collection. Section 4 presents the detailed results of the survey, including the background information of the respondents, the UML use, and the use of the MDE/MDA approaches. Section 5 discusses the research findings. Section 6 brings the conclusion, remarks, and future work.