A comparison of indirect and optimal routing in mobile networks

Abstract

When a mobile user is away from its home network, it is desired that the data transfer between the mobile user and its correspondent node could use the normal IP routing to establish routing paths between them by sophisticated routing algorithms according to the optimality principle. Unfortunately, in most of cases, the data routing under current mobile networking protocols is not so efficient as normal IP routing because packets have to be routed passing through the mobile user's home network. Here, we refer to the former as direct (optimal) routing and the later as indirect routing. Some mechanisms are suggested to enable the direct routing. However, they are not perfect and additional elements have to be introduced to support these mechanisms. Therefore, we would like to see how the indirect routing affects traffic performance and how much direct routing improves the performance. In this thesis, we summarise the basic concepts of mobile networking protocols and involved routing issues and make a comparison of indirect and direct routing in traffic performance. The traffic performance is examined based on TCP and UDP performance. Furthermore, the performance is examined for broadband and narrowband mobile user respectively. The former has an emphasis on the bottleneck within the public Internet while the later has a bottleneck within the local area, the subnet where the mobile node is attached. We design four topology scenarios that describe the relative difference between the indirect and direct routing in the length of routing path (number of hops). From topology 1 to 4, the relative difference is 0%, 50%, 100% and 433% respectively (see Figure 3.1.1-2). Topology 1 has an emphasis on encapsulation overhead since there is no difference between the two routings in length. The thesis describes the performance under various background traffic lo.ads put on the studied topologies. Given a certain background traffic load, traffic performance is examined by changing the radio link capacity, which is described by GPRS timeslot operation. The thesis concludes that for UDP data transmission the effectiveness of performance improvement of direct routing over indirect routing is sensitive to the relative difference between the two routings in the length of routing path while the effectiveness for TCP data transmission more depends on background traffic load - the higher the background traffic load is, more apparent the effectiveness is. Furthermore, it is shown that the location of bottleneck for data transmission has a significant impact on performance improvement. The bottleneck within wireless local area usually consumes the perf01mance improvement created by the direct routing.