Scalable QoS and QoS management models for IP networks.

摘要

Significant challenge has been introduced for IP networks as new emerging applications, such as VOID and E-commerce, require delivery quality guarantees and IP networks supports only best effort delivery. Three main Quality of Service (QoS) standards, Integrated Service (IntServ), Differentiated Service (DiffServ) and Multiple Label Switching Protocol (MPLS), have been introduced to handle QoS issues. These schemes use a management scheme to handle QoS requests, control bandwidth and monitor traffic flows.; A new QoS mathematical model is proposed in this thesis that guarantees requested QoS in which end-to-end delay and packet loss bounds are determined. This model is implemented at each domain then managed by a domain QoS manager. The proposed QoS manager establishes bandwidth reservation on infra and inters domain links for individual flows. In addition to resource reservation, two other essential resource control tasks within the model, admission control and traffic policing, are served. The proposed model and QoS manager retain the best features of IntServ and DiffServ however it has low complexity and provides end-to-end QoS (or per-flow QoS). The proposed new model has the following advantages: (1) Scalable: complexity is left at the edge router and core routers forward packets only. (2) Manageable : only one management system per back bone network is implemented to process QoS requests. (3) The combination is defined in this thesis as the Global Domain Identifier (GDI) and used for forwarding and reservation decisions. (4) No mapping either to DiffServ classes or MPLS formats.; As the next generation protocol, IPv6, is becoming important due to the growth of the Internet, the new model was simulated based on IPv6. The simulation demonstrates the model's performance with respect to the performance of DiffServ and IntServ, the predominate QoS methods in use today. IPv6 is intended to support QoS with a new field the Flow ID. The simulations implement the proposed new model using the Flow ID and traffic class fields to manage QoS in IPv6 domains.; The simulation results and the numerical experiments of the proposed QoS model show that a minimum end-to-end delay and minimum packet loss rate are achieved for higher priority traffic flows. In conclusion, a dynamic pricing system is integrated with the proposed QoS management system to test how prices can change when domain networks are congested. This pricing model uses the flow identifier, flow ID and the application source IP address, for calculating prices during network congestion. It also uses this identifier for charging customers. Excellent pricing and revenues results were achieved.