The DeeDS architecture is a distributed real-time systems architecture developed for complex real-time systems by the DeeDS project.
A major goal is to evaluate the effectiveness of certain approaches to ensuring predictability and efficiency in a dynamic and complex environment. DeeDS features time-cognizant reactive mechanisms complemented by multi-purpose predictable event monitoring and dynamic deadline scheduling with overload management, as well as support for delayed replication with eventual consistency. These functions are considered orthogonal and only a minimum of functionality included, to be able to control internal complexity. Predictability and efficiency are ensured by allowing all transactions to execute locally without delays due to disk accesses, network communication, or distributed commit processing. Further, a mix of criticality from hard to soft allows hard critical time constraints to be met at the expense of softer constraints, and contingency actions allow handling of transient overloads. To promote predictability, the critical system services, such as the event monitor and the scheduler, may execute on a service processor separate from the application-related functions.
More information about this project is available in our archives.