Middleware in a real-time distributed system should provide services to ensure that time-critical tasks meet their specified timing constraints. Further, these services should be coordinated in such a way that the algorithms used for schedulability analysis, request binding, priority assignment, overload management, etc. work together in a cohesive way. For example, if a rate-monotonic analysis is used to determine the schedulability of a set of periodic tasks, then the CORBA Priority assignment, local dispatching, queuing, client/server binding, and other scheduling decisions should be rate-based. We have developed a pair of strategized, coordinated Common Object Services for Real-Time CORBA (a Scheduling Service [1] and a Real-Time Trading Service [2]) that provide a framework into which coordinated real-time algorithms can be “plugged” to enforce consistent real-time behavior.. We have also investigated the problem space for real-time distributed scheduling and binding. From this investigation, we have devised a taxonomy that organizes various combinations of system parameters to define the scheduling and binding problem for particular applications.
The workshop presentation will describe the real-time middleware problem space taxonomy, the design of our strategized services, the coordination of a set of algorithms that plug into the services to solve a particular problem in the problem space, and the use of a model-integrated computing tool to automate the creation of the correct service implementations based upon the definition of the problem to be solved.