| 摘要 |
A unique method of operating a multiprogrammed computing system by dynamically regulating central processing unit (CPU) and/or peripheral resource access such that the parallel processing capability, i.e., thruput of the computing system, is increased and/or such that all programs efficiently share the computing system. The embodiment selected is dependent upon the complexity of the computing system and the amount of computing capability available for regulation. In one embodiment CPU and peripheral access is regulated in an unlike manner. In another embodiment the solution of a dynamic linear programming model, that describes the current resource requirements of two or more of the programs operating in the system, is used to regulate resource access. In yet another embodiment, the system is shared while minimizing the degradation of degradable resources. A more elaborate embodiment of this invention has the capability of alternating between two or more different regulating algorithms providing the computing system with an almost unlimited number of different operational modes. In addition, the mode itself can vary with time to allow the computing system to meet the requirements of a changing workload.
|
