Real-time applications are becoming increasingly popular in
distributed environments. These real-time applications range from hard real-time applications with periodic or aperiodic tasks and intertask relative timing constraints to soft real-time applications with best effort timing requirements. This paper introduces a complete system model for scheduling and dispatching hard as well as soft real-time tasks with intertask temporal dependencies in distributed environments. The model uses a dynamic time based off-line scheduler to verify the feasibility of a distributed hard real-time task set, and a parametric run-time kernel that guarantees the temporally determinate dispatching of hard real-time task instances and best effort performance for soft real-time task instances. The use of the dynamic time based scheduling, provides off-line guarantees for all the timing requirements of the hard real-time tasks while the parametric dispatching mechanism maintains a flexible run-time environment
that makes use of the slack time with a limited overhead. (Also cross-referenced as UMIACS-TR-2001-65)