The disappearing boundary between development-time and run-time

Modern software systems are increasingly embedded in an open world that is constantly evolving, because of changes in the requirements, in the surrounding environment, and in the way people interact with them. The platform itself on which software runs may change over time, as we move towards cloud computing. These changes are difficult to predict and anticipate, and their occurrence is out of control of the application developers. Because of these changes, the applications themselves need to change. Often, changes in the applications cannot be handled off-line, but require the software to self-react by adapting its behavior dynamically, to continue to ensure the desired quality of service. The big challenge in front of us is how to achieve the necessary degrees of flexibility and dynamism required by software without compromising the necessary dependability. This paper advocates that future software engineering research should focus on providing intelligent support to software at run-time, breaking today's rigid boundary between development-time and run-time. Models need to continue to live at run-time and evolve as changes occur while the software is running. To ensure dependability, analysis that the updated system models continue to satisfy the goals must be performed by continuous verification. If verification fails, suitable adjustment policies, supported by model-driven re-derivation of parts of the system, must be activated to keep the system aligned with its expected requirements. The paper presents the background that motivates this research focus, the main existing research directions, and an agenda for future work.