Adjustable Flow Control Filters, Reeective Memories, and Coprocessors as Support for Distributed Real-time Systems

Distributed real-time systems are diicult to construct such that the system behaves predictably. In addition, many of these systems must operate in complex and uncertain environments. We discuss several hardware-based ideas for moving distributed real-time systems towards predictability: adjustable ow control lters, reeective memories, and scheduling coproces-sors. We believe that these hardware based solutions should be exploited for enhancing predictability and ro-bustness, especially as hardware becomes cheaper. Adjustable ow control lters is a new idea being proposed here. While reeective memories are available oo-the-shelf, here we discuss their value in distributed scheduling and fault tolerance. Scheduling coproces-sors can reduce the cost of on-line scheduling so that dynamic rescheduling and early warning of impending missed deadlines can occur.