Model and analysis of burst packet losses and packet delays in IP networks using Markov chains

The classic problem of communication networks such as increase in packet delays and packet losses regains significance with the evolution of a new variety of communication networks and networking protocols. Packet delays in wired networks like the Internet were very well analyzed in the past. However, the packet delays in ad hoc networks were not thoroughly studied, especially, with respect to issues like the effect of medium access control (MAC) layer delays on the upper layer delays associated with the packet transmission process, the packet service-time, and packet end-to-end delay. Similarly, the packet losses in wired networks were analyzed using empirical models based on Markov chains. These models lack the ability to describe the response of an Internet router for different packet arrival patterns. For the networks experiencing frequent burst losses, it is very important to analyze the effect of packet interarrival time on such losses. This calls for a set of well-defined analytical models which can help in identifying the parameters responsible for the increase in packet delays and packet drops. Such models also extend to devise new network protocols and mechanisms, and improve the network throughput. This dissertation addresses the requirement of such models. The first part proposes an analytical model to describe packet delays in ad hoc networks and highlights the effect of MAC delays on route discovery time and node's packet service-time. The second part presents a semi-Markov process based model to analyze the wireless LANs (WLANs) and the MAC protocols such as the IEEE 802.11 Distributed Coordination Function (DCF) used in the IEEE WLANs. The third part presents an empirical model to describe the burst losses in Voice over Internet Protocol (VoIP) caused by queue overflows at the Internet routers. As an extension to this dissertation, an analytical model describing the combined effect of packet arrival-rate, packet service-rate, and the queue capacity on burst losses is also presented as future work.

Author

• Kadiyala, Murali Krishna

Identifier

hdl:10057/3930

OAI identifier oai:identifier

oai:soar.wichita.edu:10057/3930