دانلود رایگان مقاله روش مسیریابی کیفیت سرویس برای شبکه IP دسترسی مبتنی بر OSPF

Abstract

This paper presents a study of a method termed, multi-plane routing, that maximizes path diversity in IP routing and is targeted for IP access networks (AN). The motivation for the work is in the specific shortcomings of the conventional intra-domain IP routing principles such as “shortest-path” and “best-effort” when applied in IP ANs. We generalize these networks as the transit between the access routers and gateway and they range from a simple tree to meshed tree topologies. The method uses Multi-Topology OSPF standardized by the IETF and instantiates multiple OSPF installations in networks, each installation utilizing a portion of the topology in the conventional manner, i.e. routing plane (RP). Hence, all links are utilized by having at least one standard OSPF routing installation including them in the paths between access router and gateway. The method functions on extensions in routers and simple packet tagging allowing the routers to install and separate between paths of each RP. Routing is facilitated by the proposed method’s algorithms for network planning and traffic engineering. The former is called the offline algorithm rendering the optimum number of RPs in an arbitrary topology by independently setting link weights for each plane. The latter is called the online algorithm that applies a policy-based routing scheme for dynamically selecting the best RP based on the introduced QoS-aware cost function. The paper concludes by significant improvements in throughput, packet loss rate, session blocking and delays for numerous cases of topologies differing in numbers of networks nodes and degrees of meshing.

5. Conclusions

Building access networks’ routing and traffic engineering via extensions proposed in the paper have shown to enable significant improvements in path diversity compared to standard IP routing. The extensions required in IP routers running OSPF for implementing the MPR method are comparable to the alternative solutions, both in the performance and flexibility. In the IP routing, ECMP could be used for comparison as it is typically applied in some types of topologies, but it is not able to flexibly accommodate for overall path diversity in a high degree of meshing and large number of nodes in IP access networks. On the other hand, MPLS is able to achieve path diversity but the high overhead of its installation and maintenance present a strong case for finding the solutions in IP routing adaptations as proposed in the paper. The paper additionally promotes IP access networks as a natural extension of the infrastructure of the Internet, not requiring additional networking support in the scoped segment of the network that provides access to wireless terminals. In addition, network planning and traffic engineering via QoS-awareness and the algorithms that comprise the MPR method are features that would equally be needed for other networking solutions in IP access networks, e.g. MPLS already uses OSPF for LPS path computations and traffic engineering. QoS-aware MPR allows the network to maintain several independent logical topologies that can be used to balance the traffic load within the network whilst providing QoS for end users. Our method classifies new incoming sessions and routes them at the edge of the network, namely at the gateway and at the access nodes, onto the routing plane that achieves best network performance and that provides best QoS for the user. The method uses both an offline and an online process for network planning and traffic engineering respectively, and the performance issues were addressed both theoretically and by simulation. The results showed clearly our Q-MPR scheme outperforms existing strategies even with a small number of routing planes (5 for high-meshed access networks). Using Matlab and the NS-2 simulator, we compared Q-MPR against basic MPR, OSPF and the InvCap mechanisms. Total received throughput is increased by 45.2% with MPR compared to OSPF and InvCap strategies. Q-MPR, while generally blocking more sessions and using the same routing plane configuration as that of MPR, achieved the same overall throughput whilst lowering the total packet loss rate. For future work, we would like to investigate the portability of our method to other topologies and network models, e.g. wireless mesh networks, multiple gateways and various combinations of source/sink nodes in IP access network. Finally, we will investigate offering MPR as a candidate solution for SDN.