Adaptive architectures for future highly dependable, real time systems Brian Ford Peter Bull Alan Grigg Lin Guan Iain Phillips 2134/11751 https://repository.lboro.ac.uk/articles/conference_contribution/Adaptive_architectures_for_future_highly_dependable_real_time_systems/9404603 Many present-day safety-critical or mission-critical military applications are deployed using intrinsically static architectures. Often these applications are real-time systems, where late responses may cause potentially catastrophic results. Static architectures allow system developers to certify with a high degree of confidence that their systems will provide correct functionality during operation, but a more adaptive approach could provide some clear benefits. In particular, the ability to dynamically reconfigure the system at run time would give increased flexibility and performance in response to unpredictable or unplanned operating scenarios. Many current dynamic architectural approaches provide little or no features to facilitate the highly dependable, real-time performance required by critical systems. The challenge is to provide the features and benefits of dynamic architectural approaches while still achieving the required level of performance and dependability. This paper describes the early results of an ongoing research programme, part funded by the Software Systems Engineering Initiative (SSEI), aimed at developing a more adaptive software architecture for future military systems. A range of architectures with adaptive features (including object-based, agent based and publish/subscribe) are reviewed against the desirable characteristics of highly dependable systems. A publish/subscribe architecture is proposed as a potential way forward and a discussion of its advantages and disadvantages for highly dependable, real-time systems is given. 2013-02-22 13:49:44 Real-time Adaptive IMS SOA DCPS Information and Computing Sciences not elsewhere classified