Study On The Techniques For QoS Guarantee In The New Label Swicthing System

As the rapid growth of the Internet, all kinds of traffic, including real time services, go into the network. The problems existing in the traditional IPv4 architecture, for example the scarcity of network addresses, the large size of routing table, the low speed of packet forwarding and the less capabilities of provisioning QoS, can not satisfy the requirements of future Internet. For solving these problems, some new network architectures have been developed, including the new label switching system - Region Code Label Switching (RCLS).RCLS is a fast-forwarding mechanism similar to MPLS (Multiple Protocol Label Switching). However, differing from MPLS, it introduces a whole new concept, region code. The region code is a network address, which uniquely identifies some portion of the Internet, a geography district or an administration domain. Most importantly, a region code is able to illustrate the inherent hierarchical characteristics of the Internet. In this dissertation, the method for region code allocation and the techniques for providing QoS guarantees are studied. The main research achievements of this dissertation are as follows:1. Based on the analysis of the areas, populations, economies and development potential of most countries in the world and the allocation features of IPv4, IPv6 addresses and telephone numbers, a novel method for region code allocation with adaptivity and flexibility is developed in this paper, which can efficiently utilize the address space while representing the Internet hierarchy.2. On the basis of this method for region code allocation, the corresponding routing table structure, routing protocol and algorithm are all studied. Because of the hierarchical partition of the Internet, the size of routing table at core routers can be greatly reduced, and the traditional longest pre-fix matching in route lookup is also avoided. All these advantages are capable of accelerating packet forwarding, which benefits the QoS provisions in RCLS networks. The simulation results also show that differing form traditional routing protocols, the one in RCLS networks can not only shorten the period of becoming stable for routing tables when the topology of the network changes, but largely lessen link bandwidth and processor overhead for maintaining these entries.3. A novel distributed QoS routing algorithm in RCLS networks, the metric of which is bandwidth, is developed. This algorithm features the selection of output link according to the relationship between the bandwidth request and the available bandwidth on output links. Not only it keeps the merits of simplicity and low link overhead in traditional distributed QoS routing algorithms, but also can reduce resource fragments and admit more services into a network with heavy load. Moreover, the start threshold of this algorithm can shorten the path establishment delay while keeping performance. The extensive simulation results also validate its correctness and efficiency.4. The Internet Service Providers face a severe challenge that how maximum number of users can be admitted into networks with QoS guarantees, which also demonstrates the performance of RCLS networks. From this point, an algorithm for path level bandwidth allocation is proposed first according to the optimization programming theory. Then, on the basis of it and with some network factors, including link topology positions and injected traffic at edge nodes, taken into consideration, another algorithm for network level bandwidth allocation with load balance is developed too. This algorithm features quantitative route selection and adaptive bandwidth allocation, and is able to provide deterministic QoS guarantees while utilizing network bandwidth efficiently.5. The admission control algorithms aiming to delay sensitive services are studied. Based on the end-to-end delay upper bound of a user flow in DiffServ networks, the admission criterion and its method for bandwidth allocation at service class level are developed. Not only can it prevent the end-to-end delay of admitted user flows from violating the allowed upper bound, but make full use of network bandwidth. Moreover, this algorithm is very simple and possesses the scalability consistent with DiffServ.6. The EDCF scheme in IEEE 802.lie WLAN standards is improved. Differing from EDCF, the mended algorithm can deterministically select some frame accessing media and computes its contention window according its delay deadline. Therefore, the collisions among different access categories in the same WLAN station are avoided, and the frame's total access delay is guaranteed. By the means of probabilistic drop of time stringent frames, the total throughput of wireless channel is not degraded even while heavily loaded.