APPROACH TO ADAPTIVE ALLOCATION OF SHARED RESOURCES IN COMPUTER SYSTEMS

摘要

A request management subsystem is configured to establish service classes for clients that issue requests for a shared resource on a computer system. The subsystem also is configured to determine the state of the system with respect to bandwidth, current latency, frequency and voltage levels, among other characteristics. Further, the subsystem is configured to evaluate the requirements of each client with respect to latency sensitivity and required bandwidth, among other characteristics. Finally, the subsystem is configured to schedule access to shared resources, based on the priority class of each client, the demands of the application, and the state of the system. With this approach, the subsystem may enable all clients to perform optimally or, alternatively, may cause all clients to experience an equal reduction in performance.

主权项

1. A computer-implemented method for scheduling access requests for a shared resource, the method comprising:
receiving a first access request for the shared resource from a first client that is associated with a first service class; storing the first access request in a first queue that corresponds to the first service class; receiving a second access request for the shared resource from a second client that is associated with a second service class; storing the second access request in a second queue that corresponds to the second service class; determining that access requests issued by clients associated with the second service class should be serviced before access requests issued by clients associated with the first service class; and scheduling the second access request to be serviced by the shared resource before the first access request is serviced by the shared resource.