The Research Of Network Model Based On IP QoS

In the classical IP networks, the QoS(Quality of Service) provided by the networks was not given much attention. People only concerned about the ways to provide services as soon as possible. So there is no special method to insure the quality of service in the networks , and they can only provide "best effort" or "unreliable" services. The most important characteristic of those networks is that they only offer unreliable and connectionless services. Recently, the subscribers of networks increase quickly. At the same time, the domain in which the networks apply expands very fast. Stricter requirement is demanded by some modern real-time services. The classical IP networks can not meet those requirements very well. This problem will not be solved completely until the networks have some means that can guarantee QoS. In the first part of this thesis, the definition of QoS is given at the beginning , then the reasons why the classical IP networks can not insure packets transmission QoS are analyzed . Then three popular QoS network models (they are IntServ model, DiffServ model and IntServ & DiffServ integration model)are introduced separately in the second part of the thesis.The QoS guarantee mechanisms in the DiffServ model are studied in the third par of this thesis. The classification of data in the DiffServ model is analyzed firstly. On this classification base, thesis bring advance a route mean which hold many different forwarding information tables in single router , each classification data is forwarded according to different forwarding information table. This means which is validated by software emulation prove that it can provide much better QoS behavior in data transmission than classical forwarding mean. Then the deficiencies of DFE(Dorp From End) mechanism in a router's input queue are studied. After that, the thesis figure out that RED (Random Early Detection) active queue management mechanism can be used to prevent networks from congestion collapse. After analyzing different RED method, the thesis confirm that the RIO( RED IN/OUT) can protect data better than other method. This conclusion is testified by software emulation. At last of this part, the scheduling mechanisms are studied. By common scheduling mechanism , different data can not be insulated distinctly and they can not be handled fairly . if the networks wantto solve this problem, there must be some fair scheduling mechanisms in them. So this paper analyzes many fair-scheduling mechanisms. On this base, a new scheduling method is put forward. The new method proved by emulation is priority-based DRR(Deficit Round Robin) which can not only meet the requirement of the data with strict QoS restriction but also save the bandwidth. The last part is the conclusion for whole paper. DiffServ model can fulfil the request of data transmission a certain extent without changing classical IP mechanism much. On the other hand , the thesis consider that if the new route method , RIO and priority-based DRR are used at one time in a network, the transmission requirements of data with high QoS demand can be satisfied well...