The problem of placing mobility anchor points in wireless mesh networks

Reviewer: Alexandru Petrescu

The mechanisms typically used in wireless mesh networks (WMNs) consider the random topology of nodes as a fundamental design requirement; for example, mobile ad hoc network (MANET) protocols such as radio-aware optimized link-state routing (RA-OLSR) assume arbitrary graph topologies. This paper, however, considers the use of hierarchical mobile Internet protocol, version 6 (HMIPv6)-a protocol designed for inherently tree-like structures-deployed on a random WMN layout. This obviously poses a problem when planning the placement of the mobility anchor point (MAP)-a particular router in HMIPv6 terminology-in the nodes of a WMN. The method should respect the following requirements: not every node can be a MAP, the handover latency of a mobile node should be minimized, and use as few MAPs as possible. To solve the problem, Wu and Landfeldt start by qualifying its most generic version as being nondeterministic polynomial time (NP)-hard. Then, integer programs are specified as a "minimize" mathematical expression coupled to a similar "subject to" expression. This is considered for a chain topology, for a grid, and, finally, for an arbitrary topology. A greedy placement algorithm is given. From these, OMNet++ simulations are performed. The results are graphed: average handoff latency, handoff latency versus latency to the home agent (HA), number of stations against handoff latency, and number of stations against packet loss rate. As with most ad hoc network simulation literature, the paper is mostly an academic exercise that applies a conceptual model to an image of what a network could be. It does not point to any implementation and the figures given are not related to physical phenomena. It is an interesting informal exercise. Online Computing Reviews Service