Queuing analysis of dynamic resource allocation for virtual routers

摘要

The most critical issue in network virtualization is the dynamic resource allocation of the physical substrate. There is a need to monitor the running virtual routers in order to allow an adaptive change in the resource allocation. In this paper, we focus on the router data plane virtualization and we explore this issue by presenting a new dynamic allocation approach through queuing theory. We consider the problem where multiple instances of virtual routers (VRs) that have some quality of service (QoS) requirements are sharing different physical resources. We propose a novel router architecture that offers a strong isolation between concurrent VRs and provides a dynamic allocation scheme in order to guarantee the provided QoS to each of them. Our approach aims at providing a higher isolation for the concurrent virtual routers sharing the same infrastructure. We propose a dynamic Weighted Fair Queuing (WFQ) scheduler for each physical resource and an algorithm for adjusting the weight of each VR in order to reduce the delay of the packet processing and avoid the bottlenecks. We also propose an admission control mechanism that estimates the current load of the physical node and decides either to accept or reject a creation request of a new VR. Simulation results show that the proposed approach achieves good performance in terms of delay minimization inside the virtual router.