A Comparative Study of Two Generic Resource Allocation Models

In a typical resource allocation model, clients request a fixed amount of a certain resource from the resource manager (RM). Based on the current allocations, the RM grants or rejects the request. An alternative abstract model for the resource allocation problem assumes that the client specifies his request as a range of possible allocations rather than a fixed point requirement. The client is assumed to have the ability to dynamically adapt to the resource level (QoS) provided by the RM, which can change over time. Each request has a priority associated with it, and the objective of the RM is to maximize the allocation for all the clients while refusing the minimum number of requests. Both models can be either non-preemptive or preemptive based on priorities. In the range-based model, degradation of the level of resource allocation is not considered as preemption as long as the minimum requirement of the client is satisfied. In this paper we propose two methods for allocating resources using the range-based model. We propose a unified metric for judging the effectiveness of resource allocation, and we compare the performance of the proposed methods and contrast it to the performance of fixed point based resource allocation, in both the preemptive and the non-preemptive cases.