Resource sharing with sliding constraints
View Patent ↗The resource sharing system according to the present invention dynamically adjusts the priorities at which requests from applications in different request classes (or classes of service) for a shared resource, are processed. The dynamic priority of a request class is based in part on the average resource allocation to requests in this request class, and in part on settings for the minimum and maximum allocations to the request class. The average resource allocation is the proportion of time the shared resource has been assigned to requests of this class relative to other classes. The dynamic adjustment of priorities uses “sliding constraints” whereby the priorities are caused to “slide” with the average resource allocation, and the priority imposes a constraint on when the requests of a resource class can be satisfied: namely, when there are no other requests from higher-priority request classes waiting for the resource.
1. A computer implemented method of sharing a resource amongst a plurality of applications issuing requests in different request classes, comprising the steps of:
i) dynamically assigning a priority to each of a plurality of request queues associated with respective ones of said request classes, such assigning being in accordance with a moving average resource allocation to each of said respective request classes and the prionty assigned to a respective request queue being a function of the moving average resource allocation to the associated request class, wherein,
p i =1.0 if u i <=min i
p i =1.0−( u i −min i )/(max i −min i ) if min i< u i <=max i
p i =0.0 if u i >max i
where P i is the priority assigned to the i th one of said queues associate with the i th one of said request classes, and u i is the moving average allocated to said i th one of said request classes and min i and max i are respectively the minimum and maximum allocations to said i th one of said request classes;
ii) receiving and queuing said requests from said applications in said plurality of queues in accordance with said respective request classes;
iii) allocating said resource to one of said applications whose request has been queued longest in a highest priority one of said queues; and
iv) in response to said one of said applications relinquishing said resource, repeating steps i) to iii).
2. The computer implemented method of claim 1 , further including the step of updating said moving average of each of said request classes immediately upon allocating said resource.