| Summary: | 碩士 === 國立中央大學 === 資訊工程研究所 === 87 === The concept of virtual private networks (VPN) is to separate logically several private networks from the public networks by using the technology of tunneling. Besides, security and quality of services (QoS) are supported for VPNs. The bandwidth allocated to VPN tunnels may be wasted for low utilization, and the offered number of users for the system is thus decreased. On the contrary, if the bandwidth of VPN tunnels is less-allocated to increase the utilization, but, the blocking probability will be increased. Therefore, it is important to design a dynamic bandwidth allocation scheme for virtual private networks over public networks.
A two-layer dynamic bandwidth allocation scheme is proposed to assign bandwidth for VPNs, where the control architecture is divided in VPN layer and call layer. While a new VPN is created VPN layer control mechanism allocates the predefined low bound bandwidth to the VPN. The bandwidth assigned may be increased only if the assigned bandwidth is not enough. The bandwidth allocation in the call layer is to manage bandwidth already assigned for a VPN. Although the residual bandwidth of the link is less than the low bound, a tunnel with the residual bandwidth is still to be created temporarily, and the tunnel is called to be staying in temporary state. The bandwidth of the tunnel will be increased to the low bound as long as there is released bandwidth. The network will offer more VPNs by introducing the concept of temporary state, and the link utilization is thus improved. Simulation results show blocking probability can be decreased by 11.92% ~41.87%.
For the call layer bandwidth allocation, there are three schemes for bandwidth increase of tunnels: (1) increase the bandwidth by a call, (2) increase the bandwidth by a step (several calls), and (3) increase the bandwidth by a ratio. There are also three schemes for bandwidth decrease of tunnels: (1) release the bandwidth immediately as long as a call departs, (2) release the bandwidth when the residual bandwidth of the tunnel is larger than a step, and (3) release the bandwidth when the residual bandwidth of the tunnel is larger than a ratio of the current allocated bandwidth of the VPN. It is found that the policy using the third bandwidth increase and the third bandwidth decrease schemes is capable of decreasing the call blocking probability by 25.74% VPN and NPL by 95.51%
|
|---|
