A simulation approach to the design of dynamic feedback scheduling algorithms for time-shared computer systems

The goal of a scheduling algorithm for a time-shared computer system is to provide acceptable request response time and resource utilization through effective resource allocation. In order to do this, it is necessary for the algorithm to be capable of adjusting itself to handle the various situations, precipitated by the set of active user requests and the computing system's status, which may occur. An effort is now underway to design the structural framework of a scheduling algorithm which will dynamically formulate its resource allocation policies and adjust its policy formulation depending on the success or failure of those policies. Once designed, the framework will then be used to construct the scheduling algorithm for a given time-shared computer system.

The approach chosen for determining the practicability of the algorithm design is the inclusion of the algorithm in the simulation of a swapping time-shared computer system model. The simulation contains four basic activities: processor, input/output, swapper, each of which contains a resource allocation policy for determining a priority ordering among those requests which have asked for the respective resource, and the user activity. User activities follow one of a number of simulated scripts which may be composed of several types of requests. Each request is given behavioral traits dependent on its type.

While the complete algorithm is not yet implemented within the simulation, preliminary results suggest that some improvement in response time may be possible.