Rate-based scheduling of multiplexed resources.

摘要

The invention comprises method and apparatus for time-multiplexing a resource (MR) among entities which require access to the resource at a rate which is controlled over both an averaging interval and a jitter averaging interval which is shorter than the averaging interval. The technique employs a set of frame queues (117) of work descriptors (103) representing the work to be done by the resource for the entities. Each frame queue is associated with a frame of rime slots in which the resource performs work. A resource server (121) provides work descriptors from the head of a given frame queue for the frame associated with the frame queue and then serves the next frame queue for its frame. A queue manager (107) places work descriptors received from the entities at the tails of the frame queues. The work descriptors from a given one of the entities are placed in the frame queues in such fashion that when the resource server takes work descriptors from the heads of the frame queues, it automatically provides the work descriptors for a given entity to the resource at the required controlled rate. The resource server may serve one or more best effort queues (115) in addition to the frame queues. The work descriptors (103) in the best effort queues represent work for entities which do not required a controlled rate of access. The resource server serves a best effort queue when a frame queue is empty but there are time slots remaining in its frame. <IMAGE>