The Fading Boundary between Development Time and Run Time

Summary form only given. Modern software applications are often embedded in highly dynamic contexts. Changes may occur in the requirements, in the behavior of the environment in which the application is embedded, in the usage profiles that characterize interactive aspects. Changes are difficult to predict and anticipate, and are out of control of the application. Their occurrence, however, may be disruptive, and therefore the software must also change accordingly. In many cases, changes to the software cannot be handled off-line, but require the software to self react by adapting its behavior dynamically, in order to continue to ensure the required quality of service. The big challenge in front of us is how to achieve the necessary degrees of flexibility and dynamism required in this setting without compromising dependability of the applications. To achieve dependability, a software engineering paradigm shift is needed. The traditional focus on quality, verification, models, and model transformations must extend from development time to run time. Not only software development environments (SDEs) are important for the software engineer to develop better software. Feature-full Software Run-time Environments (SREs) are also key. SREs must be populated by a wealth of functionalities that support on-line monitoring of the environment, inferring significant changes through machine learning methods, keeping models alive and updating them accordingly, reasoning on models about requirements satisfaction after changes occur, and triggering model-driven self-adaptive reactions, if necessary. In essence, self adaptation must be grounded on the firm foundations provided by formal methods and tools in a seamless SDE SRE setting. The talk discusses these concepts by focusing on non-functional requirements-reliability and performance-that can be expressed in quantitative probabilistic requirements. In particular, it shows how probabilistic model checking can help reasoning about re- - quirements satisfaction and how it can be made run-time efficient. The talk reports on some results of research developed within the SMScom project, funded by the European Commission, Programme IDEAS-ERC, Project 227977 (http://www.erc-smscom.org/).