Study On Packet Switching Technology Supporting QoS Based On Universai Network

With the rise of multimedia services in information networks, structure of routing switching equipment can or not provide QoS guarantee is concerned again, how to achieve switch technologies with QoS guarantee have received extensive attention in the academia and industry. The current research about switch structure and algorithm are mainly concentrated on how to improve throughput, reduce latency and so on. Along with more different applications increasing in the future, the switch structure must be able to provide more abundant QoS guarantee. The existing information network's original design idea is essentially one kind of network supports one kind of main service, the evolution and development based on it are difficult to break the limitation of original design idea, unable to satisfy the multiple demands of network and service."Universal network", by three times mapping, make the identify theory through the entire network structure from the service layer to the network communication layer, different service types can be reflected to network communication layer through the mapping. Combination with the research of national 973 plan project"Architectural model and exchange routing theory and techniques of Universal network"in the 11th five-year plan, against the QoS demand of routing switching equipment, analysis the Universal network structure, the insufficiency in QoS guarantee of existing typical switch structures and algorithms in-depth, we firstly introduce the identifier to switch structure and propose the switch technology support QoS based on it in this paper. The switch based on identifier can avoid the depth analysis about data packet and carry on the recognition to different traffic classes.The main research work of this paper includes:1. Analysis and comments the traditional typical switch technology; propose the switch technology based on identifier. From the QoS guarantee of switch system, comprehensively classify, analysis and compare for the current switch structure and algorithm, traditional switch are based on each of the flow. Because at the same time there are may as many as a million data flows through the router, so the routing, switching, and QoS control use data flow as a unit is extremely difficult to achieve under the existing hardware conditions, if adopt Diffserv technology and use aggregate flow as a unit to management data flow, it is easily, but the granularity is too coarse, only 8 grades, it can only provide fewer QoS guarantee. We introduce the identifier to switch structure and propose the switch technology based on it suitable for universal network, and give the features of the switch technology should have. The switch technology management the network data flow in denomination of traffic classes, the number of traffic class is basically hundreds of orders of magnitude, which is not only easy to manage, but also reflect the QoS requirements of different traffic classes accurately.2. Studied the QoS supported CICQ mechanism based on identifier. Give the definition and explanation of identifier and differentiated QoS used in communications. Design QoS guarantee switch structure DS-CICQ based on CICQ and proposed a distributed dynamic dual round-robin scheduling algorithm ID-DDRR for DS-CICQ. Between different type of business and output ports, adopt double pointer double round-robin scheduling mechanism, the complexity reduces greatly. The quantum function is based on queue length and cell priority. ID-DDRR algorithm can provide different level QoS guarantee for date packets, can achieve all reservation bandwidth allocated between GD, GB and LG classes fairly. Theoretical analysis and simulation consistently show that ID-DDRR has good latency and fair performance, compare with the exiting algorithm, can support differentiate service better and realizes QoS guarantee.3. Propose a new differentiated QoS support solutions PSVIOQ-CICQ based on identifier. Analysis the main research about PPS switch technology, propose a new PPS structure—PSVIOQ-CICQ based on CICQ.The design of PSVIOQ-CICQ introduce virtual queue technology, input buffer is set to N VOQ, collaborate with the load balancer complete the balanced allocate of load. Output buffer is set to N VIQ, collaborate with the cell integrator complete the cells forward in order and provide QoS guarantee. Simulation shows that PSVIOQ-CICQ has good fairness, load balancing and scalability performance, achieve the design objects overall. At the same time, simulation shows that the throughput and delay performance of it under non-uniform traffic source are insufficient, especially when the load is heavier.4. Against the shortage of PSVIOQ-CICQ, we proposed another solution PSCICQ, in each demultiplexer, introduce NK cells size of buffer, and propose a differentiated QoS support scheduling mechanism for PSCICQ with identifier. Proved PSCICQ can guarantee cells in order. Using the distributed scheduling features of buffered crossbar, management the network data flow in denomination of traffic classes, set a small amount of buffer in the multiplexer and adopt double pointer round-robin algorithm DPRR to achieve differentiated QoS guarantee, ensured the switch provide effective support for different high-level traffic classes. Theoretical simulation result show that the new differentiated QoS support scheme PPS is effective, the throughput of it can reach up to 99% at full load, can according to the reservation bandwidth to allocate the bandwidth of output links, ensure packets have an upper delay bound, can allocate load to middle switch plan balanced, has good equilibrium degree of load distribution, compared to the current mainstream PPS design, the mechanism easily implemented in hardware, has better extensibility.5. Studied the QoS support switch structure performance evaluation model based on utility function. Analysis the study status of utility function about traffic flows, given the latency utility function for traffic flows, propose a new switch performance evaluation model called"based on latency utility function", and with utility maximization theory to find out the optimum allocation scheme of network resources. And then we extend the model, integrate the two most closely to the nature of traffic flow and user-intuitive feel indicator, the latency and bandwidth, into our model, making the evaluation more comprehensive, intuitive and practical. That is the double-indicator evaluation model. The scheme can able to use the exiting network resources effectively, and provide better QoS for every kind of traffic flow as far as possible.