A sharing-oriented service selection and scheduling approach for the optimization of resource utilization

Abstract

We present a sharing-oriented service selection and scheduling approach capable of finding a trade-off between requirement satisfaction degree, service utilization rate and service sharing cost for limited quantities and capacities of available services. In traditional service selection approaches, each customer requirement is independently satisfied by optimally selecting a set of candidate service resources. However, in real-life service scenarios, it is usual for multiple customers to raise their requirements simultaneously, and available services need to be allocated between them. Especially, when available services are limited in both quantity and capacity, a traditional "first-come-first-serve" strategy would lead to a low service utilization rate, and some requirements cannot be satisfied at all (i.e., a low requirement satisfaction degree). Our approach makes use of the feature that some services can be shared by several customer requirements. Specifically, a virtualized service resource consisting of multiple candidate services is constructed and scheduled to satisfy multiple customer requirements simultaneously. Our approach searches for the global optimization on requirement satisfaction degree, service utilization rate, and service sharing cost. We build a mathematical model for this multi-objective optimization problem and propose a nested genetic algorithm mixed with a greedy strategy. Experiments in an ocean transportation service setting are conducted and our approach is compared with traditional approaches to validate its effectiveness.