Virtual network embedding with multiple priority classes sharing substrate resources

Abstract

Network virtualization is a promising approach for virtual network operators to configure their own inter-cloud networks flexibly on an inter-cloud substrate network. Virtual network operators specify multiple priority classes in order to satisfy different latency requirements for their inter-cloud network services with a variety of delay-sensitivities on their virtual networks. Furthermore, they may require substrate resource sharing among multiple priority classes in order to reduce the amount of substrate resources assigned to them. Meanwhile, it is desirable for a substrate network provider that multiple virtual networks can share substrate resources since the total amount of substrate resources required can be reduced due to the effect of statistical multiplexing. This paper formulates a novel virtual network embedding problem and proposes a heuristic virtual network embedding method to minimize the total substrate resources required when the virtual network operators request substrate resource sharing among multiple priority classes within their virtual networks. Based on the proposed method, multiple virtual networks can maximally share substrate resources with one another while sharing substrate resources on an equal basis. The effect of substrate resource sharing among multiple priority classes and multiple virtual networks is quantitatively assessed through extensive simulations, and advantageous topologies for requested virtual networks and substrate networks are also presented.