VPN Provisioning over WDM Networks

碩士 === 國立臺灣大學 === 資訊管理研究所 === 90 === The emergence of virtual private networks (VPN) provides an alternative choice for people or companies who want to have their own networks. It not only substantially reduces the cost, but also speeds up the construction of private networks. In the operators’ poin...

Full description

Bibliographic Details
Main Authors: Hung Tien Chung, 鐘泓田
Other Authors: Yeong-Sun Lin
Format: Others
Language:zh-TW
Published: 2002
Online Access:http://ndltd.ncl.edu.tw/handle/82369167267171220321
Description
Summary:碩士 === 國立臺灣大學 === 資訊管理研究所 === 90 === The emergence of virtual private networks (VPN) provides an alternative choice for people or companies who want to have their own networks. It not only substantially reduces the cost, but also speeds up the construction of private networks. In the operators’ point of view, admitting new VPNs can bring them additional revenue, but that’ll also consume some resources of the limited physical network and even influences the transmission quality of the existing users. In this thesis, we want to develop an admission control algorithm for VPNs over WDM networks, where the objective function is to maximize the total revenue, subject to various QoS requirements as agreed upon in the SLAs signed with the VPN users. The whole design problem can be further decomposed into the following three subproblems: 1.VPN admission control subproblem: to determine which VPN can be admitted into the network. 2.IP layer routing subproblem: to decide an optimal path for an given OD pair. 3.WDM layer lightpath assignment subproblem: to construct and decide the wavelength assignment of lightpaths for each IP link. We develop a mathematical programming model to describe the joint design problem. It turns out to be a non-linear non-convex mixed integer programming problem which becomes more difficult and complicated for jointly considering many decision variables at the same time. A number of heuristic solutions based on Lagrangean relaxation methods are proposed to solve the complicated problem, which solve the above three subproblems simultaneously. According to the computational experiments, the Lagrangean relaxation based algorithms achieve up better performance than that of other primal based heuristics.