The adoption of Agile Software Development in large companies is a recent phenomenon of great interest both for researchers and practitioners. Although intra-team interaction is well supported by established agile practices, the critical interaction between the agile team and other parts of the organization is still unexplored in literature. Such interactions slow down the development, hindering the achievement of business goals based on speed: short time to market, quick replication of products of a product-line, and reaction time for product evolution. We have employed a two-year long multiple-case case-study, collecting data through interviews and a survey in three large companies developing embedded software. Through a combination of qualitative and quantitative analysis, we have found strong evidence that interaction challenges between the development team and other groups in the organization hinder speed and are widespread in the organizations. This paper also identifies current practices in use at the studied companies and provides detailed guidelines for novel solutions in the investigated domain. Such practices are called boundary-spanning activities in information system research and coordination theory. We present a comparison between large embedded software companies employing agile and developing a line of products based on reused assets and agile companies developing pure software. We highlight specific contextual factors and areas where novel spanning activities are needed for mitigating the interaction challenges hindering speed. Copyright © 2015 John Wiley & Sons, Ltd.

Model Driven Engineering (MDE) is a proven approach to improve software development processes by automation. However, traditional development of MDE tooling requires a high upfront cost. Recent developments in language workbench technologies promise to significantly reduce these investment costs. By providing domain experts with targeted projections, the speed and quality of delivering customer value is improved. This paper provides results from an industrial case study in the telecommunications domain and compares the value of using a language workbench to traditional MDE technologies. Evaluation of the approach was based on qualitative research strategy which involved a proof of concept implementation and effort estimations by tooling experts. Our results, using the Intentional Domain Workbench, indicate that applying a language workbench promises significant improvements in several aspects of MDE based software development. Most notably in this paper: (1) improved speed in development of domain specific tooling and (2) improved speed in software development process re-engineering.

An open issue in industry is the combination of software reuse in the context of large scale Agile Software Development. The speed offered by Agile Software Development is needed for short time to market, while reuse strategies such as Software Product Line Engineering are needed for long-term productivity, efficiency, and profit. The paper investigates, through a survey, communication factors affecting both speed and reuse in 3 large companies developing embedded systems and employing Agile Software Development and Software Product Line Engineering. Our results include a prioritized list of communication related factors obtained by statistical analysis and the recognition and spread of the factors in the companies. We have recognized 5 interfaces with the Agile development team that need to be improved: system engineers (architects), product management, distributed teams, inter-project teams and sales unit. Few factors (involving inter-project communication) depend on the business drivers for the company. We also reveal that Agile teams need strategic and architectural inputs in order to be implanted in a large company employing Software Product Line Engineering. Academic and industrial training as well as different tactics for co-location would improve the communication skills of engineers. There is also a need for solutions, in the reference architecture, for fostering Agile Software Development: the goal is the combination of the focus on customer value of the teams, reusability, system requirements and avoidance of organizational dependencies.

An open issue in industry is software reuse in the context of large scale Agile product development. The speed offered by agile practices is needed to hit the market, while reuse is needed for long-term productivity, efficiency, and profit. The paper presents an empirical investigation of factors influencing speed and reuse in three large product developing organizations seeking to implement Agile practices. The paper identifies, through a multiple case study with 3 organizations, 114 business-, process-, organizational-, architecture-, knowledge- and communication factors with positive or negative influences on reuse, speed or both. Contributions are a categorized inventory of influencing factors, a display for organizing factors for the purpose of process improvement work, and a list of key improvement areas to address when implementing reuse in organizations striving to become more Agile. Categories identified include good factors with positive influences on reuse or speed, harmful factors with negative influences, and complex factors involving inverse or ambiguous relationships. Key improvement areas in the studied organizations are intra-organizational communication practices, reuse awareness and practices, architectural integration and variability management. Results are intended to support process improvement work in the direction of Agile product development. Feedback on results from the studied organizations has been that the inventory captures current situations, and is useful for software process improvement work.

Efficient architecture work involves balancing the degree of architectural documentation with attention to needs, costs, agility and other factors. This paper presents a method for prioritizing architectural concerns in the presence of heterogeneous stakeholder groups in large organizations that need to evolve existing architecture. The method involves enquiry, analysis, and deliberation using collaborative and analytical techniques. Method outcomes are action principles directed to managers and improvement advice directed to architects along with evidence for recommendations made. The method results from 3 years of action research at Ericsson AB with the purpose of adding missing views to architectural documentation and removing superfluous ones. It is illustrated on a case where 29 senior engineers and managers within Ericsson prioritized 37 architectural concerns areas to arrive at 8 action principles, 5 prioritized improvement areas, and 24 improvement suggestions. Feedback from the organization is that the method has been effective in prioritizing architectural concerns, that data collection and analysis is more extensive compared to traditional prioritization practices, but that extensive analysis seems inevitable in architecture improvement work.

Requirements modeling is an established method for detecting defects in requirement specifications. Although many companies have successfully incorporated requirements modeling in their software processes, other companies face technology adoption problems: methods appear incompatible with processes and technologies in use, and with the attitudes of engineers. We approach this problem by adaptation of standard modeling know-how to existing practices and attitudes in certain companies. By a requirements modeling process that respects the constraints of iterative, incremental, use case driven development, and that uses a modeling language with a low learning barrier (for engineers), we hope to make companies that use the unified process amenable to requirements modeling.

We present a functional language with a type system such that well typed programs run within stated space-bounds. The language is a strict, first-order variant of ML with constructs for explicit storage management. The type system is a variant of Tofte and Talpin's region inference system to which the notion of sized types, of Hughes, Pareto and Sabry, has been added.

We have designed and implemented a type-based analysis for proving some basic properties of reactive systems. The analysis manipulates rich type expressions that contain information about the sizes of recursively defined data structures. Sized types are useful for detecting deadlocks, nontermination, and other errors in embedded programs. To establish the soundness of the analysis we have developed an appropriate semantic model of sized types.