Performance Studies of Layered MIPv6 based Network Architecture for better QoS: a Mathematical Approach

摘要

Layered architecture is considered as a suitable solution for seamless mobility for Mobile IPv6 based network in terms of handoff latency. But too many layers in the hierarchy introduce large amount of packet drop at the anchor agents and degrade the efficiency of anchor agents. So, determining the optimal layers of hierarchy for better QoS parameters such as handoff latency, packet loss and overall efficiency is a challenging research area. The work presented in this paper is a mathematical analysis of a general n-layered architecture to find optimal levels of hierarchy with minimum handoff latency and packet dropping probability along with a measure of efficiency of anchor agents at different layers. The factors that affect the performance of the hierarchical model are the number of mobile nodes under an anchor agent and packet arrival rate at the agent. We assume that each of the anchor agents in the network maintains an M/M/1/K queue and the packet arrival rate at the anchor agent follows Poisson's distribution. Analysis shows that handoff latency decreases by a ratio of 25-35% with the addition of a new layer up to layer three, around 15% decrease on adding fourth layer and a negligible decrease of 2-3% beyond layer four. Also, packet-dropping probability is directly proportional to the offered load, which in turn is dependent on the number of mobile nodes. As the number of layer increases, the coverage area of the anchor agent as well as the mobile nodes under its coverage increases. A 2-5% of the packets are dropped up to layer four beyond which it exceeds 5%. A 5% handoff dropping is not considered to be acceptable. Also, the efficiency of anchor agent remains above 97% up to layer three for most of the packet arrival patterns and MN density. So, considering the performance parameters such as handoff latency, packet dropping probability and efficiency of anchor agent, a three-layered architecture may be considered optimal.