Philippe Kruchten

Release planning is the decision process by which we decide what we are going to do next in a software project, for the best of its stakeholders, and in face of all the uncertainties inherent to such endeavour. While a simple strategy can focus simply on ‘delivering value’, this rapidly becomes much more complicated when we must take into consideration dependencies, value and cost, software architecture, existing defects, and the dark spectre of technical debt. In this presentation, I introduce a gradually more and more sophisticated strategy, drawing on various established techniques to deal with this delicate balancing act.

Philippe Kruchten is first and foremost a professional software engineer, with thirty years of industrial experience developing large-scale software-intensive systems in the domains of telecommunications, aerospace, defense, transportation and software tools. From 1996, he was director of process development at Rational Software, where he developed the Rational Unified Process (RUP). He kept his position when Rational was acquired by IBM in 2003. Since 2004, he is professor of software engineering at the University of British Columbia in Vancouver, Canada. Kruchten has authored numerous publications, of which ‘The 4+1 view model of architecture’ is one of the most cited articles in the field of software engineering.

Tim Trew

When developing a new product line, what decisions have to be made at the architectural level, and which can be left to subsystem developers? What happens during integration when, despite a top-level decomposition and good interface specifications, there is insufficient shared vision across development teams? And why are the consequences of overlooked decisions much worse for a product line, compared with an evolving single system?

We have developed a reference architecture for a broad domain of embedded products, ranging from TVs to medical imaging systems, that captures the lessons learnt from problems encountered in early product line developments, and presents decisions in a form that allows architects to apply them in their own context. This is particularly challenging for embedded products, where the variety of hardware architectures, and the changing allocation of functionality between hardware and software, limits the reusability of concrete product architectures. More abstract approaches, such as pattern languages and architectural knowledge repositories require the architect to understand an extensive conceptual framework before they can be used.

The presentation will describe the development of a reference architecture that captures architectural best practices from a wide variety of embedded products and packages them in a form that is easy to reapply.

Until recently, Tim Trew was a research fellow at NXP Semiconductors (previously Philips Semiconductors), Eindhoven. He has focussed on software integration issues since 1995, identifying the causes of integration failures, how to test for them and the design approaches that should be adopted to eliminate them. This work was informed by his involvement in many ‘first-of-a-kind’ developments in Philips. This led him to work on the creation of guidelines for the development of architectures, the assessment of the technical risks associated with integrating third-party software and the challenges of developing embedded software in supply chains.