Research On Network Service Selection And Network Resource Embedding In Software Defined WAN

With the diversification of network applications and services,the existing network architecture and service mode cannot meet the needs of network transiting required by different users or applications.Many players in the WAN ecosystem,such as network operators,content providers,infrastructure providers have their own interests.The emergence of Software Defined Networking and Network Virtualization brings about a good opportunity of thorough innovation.The rise of the Software Defined WAN makes it possible to establish one more effective and more direct relationship between users and ISPs.Against this background,the “Network on Demand” launched by AT&T and China Telecom attracts users to custom the network and select appropriate network service for themselves.Meanwhile,ISPs can reconstruct and integrate resources and provide service for maximum utilization.Given the deployment of more application services in the cloud environment,the bandwidth of wide area network and data centers has become a vital network resource.ISPs update existing pricing strategies to address the rapidly changing needs of services and applications for bandwidth.ISPs pass the cost of pressure on to the users,resulting in gradual increase of the cost of renting the Internet resources.At the same time,in many situations,service demand in wide area network is difficult to predict and describe by users,which usually drives them to take a conservative approach,renting large bandwidth to guarantee business functioning,and fuels operators to oversell the network resource.In addition,many of the applications of the network have a higher demand for delay,such as instant messaging,video conferencing,search engines,etc.When constructing a user-customized virtual network,there is a lack of a latency-awared mapping method between the virtual network and the physical network.In view of the above situation,from the users' point of view,this paper establishes network service selection model under uncertain circumstances of users' demands to predict the users' short-term demands,as well as guides the users to select the most appropriate network services while the ISPs provides various pricing schemes.Meanwhile,the proposed time slot based users' demand cost model utilizes the Lyapunov optimization technology to help ensure that users can minimize the operation costs and achieve the goal of guaranteeing the deadline constraint of transmission and capacity constraint of links over the whole network.Then,from the view of ISPs,under the condition of accessing users' demands,proposed latency-awared vSDN controller embedding algorithm targets at guaranteeing the latency between the controllers and virtual switches,enhances the performance of virtual network embedding and promotes their competitiveness by improving the level of service ability.Finally,through simulation and in accordance with the topology of CERNET,we construct the experiment environment.Compared with the existing minimum delay algorithm,the proposed algorithm in this paper increases the number of the accepted virtual network requests,while assuring the users' the demands of latency.The index mentioned is bound up with the revenues of the ISPs,verifying the validity and correctness of the proposed algorithm.