Due to the high availability of cheap hardware, wireless multi-hop networks and in particular Wireless Mesh Networks (WMNs) are becoming popular in more and more contexts. For instance, IEEE 802.11 based WMNs have already started to be deployed as means to provide Internet access to rural areas in the developing world. To lower the cost and increase the coverage in such deployments, the wired network is extended with a wireless backbone of fixed mesh routers. With advances in technology and reduction in price comes also the possibility for more powerful wireless nodes, having multiple radios that allow transmitting on different channels in parallel. To be a successful platform for providing general Internet access, wireless multi-hop networks must provide support for common Internet applications. As most of the applications in the Internet today use the Transmission Control Protocol (TCP), TCP performance is crucial. Unfortunately, the design of TCP’s congestion control that made it successful in today’s Internet makes it perform less than optimal in wireless multi-hop networks. This is due to, among others, TCP’s inability to distinguish wireless losses from congestion losses. The current trend for operating system designers is also to focus TCP development on high-speed fixed networks, rather than on wireless multi-hop networks. To enable wireless multi hop networks as a successful platform there is therefore a need to provide good performance using TCP variants commonly deployed in the Internet. In this thesis, we develop novel proposals for the network layer in wireless multi-hop networks to support TCP traffic more efficiently. As an initial study, we experimentally evaluate different TCP variants, with and without mobile nodes, in a MANET context. Our results show that TCP Vegas, which does not provoke packet loss to determine available bandwidth, reduces the stress on the network while still providing the same or slightly increased performance, compared to TCP Newreno. We further propose and evaluate packet aggregation combined with aggregation aware multi-path forwarding to better utilize the available bandwidth. IP layer packet aggregation, where small packets are combined to larger ones before sent to the link layer, has been shown to improve the performance in wireless multi-hop networks for UDP and small packet transfers. Only few studies have been made on the impact of packet aggregation on TCP traffic, despite the fact that TCP traffic constitutes the majority of the Internet traffic. We propose a novel aggregation algorithm that is specifically addressing TCP relevant issues like packet reordering, fairness and TCP timeouts. In a typical WMN scenario, the aggregation algorithm increases TCP performance by up to 70 % and decreases round trip time (RTT) by up to 40 %. A detailed evaluation of packet aggregation in a multi radio setting has shown that a naive combination of multi path routing and packet aggregation can cause valuable aggregation opportunities to be lost. Therefore, we propose a novel combined packet aggregation and aggregation aware forwarding strategy that can reduce delay, packet loss and increase TCP performance by around 30 %.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:kau-6820 |
| Date | January 2011 |
| Creators | Karlsson, Jonas |
| Publisher | Karlstads universitet, Avdelningen för datavetenskap, Karlstad : Karlstad University |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Licentiate thesis, comprehensive summary, info:eu-repo/semantics/masterThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | Karlstad University Studies, 1403-8099 ; 2011:5 |
Page generated in 0.0026 seconds