On the problem of placing Mobility Anchor Points in Wireless Mesh Networks

1. On the problem of placing Mobility Anchor Points in Wireless Mesh Networks Lei Wu & Bjorn Lanfeldt, Wireless Mesh Community Networks Workshop, 2006 므리나 September 18, 2007

2. Outline • Introduction • Problem description • Related work • Analysis of the problem in mathematical mode • Identification of the most important factors • Proposed approach • Simulation validation • Future work & conclusion CS 712

3. Introduction • Wireless mesh networks (WMN) • Multi-hop wireless networks • Consist of wireless routers and mesh clients • Mobility of mesh clients in WMN • Low mobility as compared to mobile ad-hoc networks • WMN Vs current fixed mobile networks • Unplanned topology • Wireless connection between mesh routers • Two geographically neighboring nodes can be very far from each other CS 712

4. A Quick Overview • Mobile IP • MN moves from a network coverage cell to another cell • Gets a CoA from the visited network • Registers the association between the CoA and the HA by sending a binding message • Hierarchical Mobile IP • Localized mobility management style to reduce the signaling overhead • Manages local and global mobility separately • To deal with local movement management • MAP is introduced • Entity that deals with the MIP handoffs • Local and global binding updates CS 712

5. Problem Description • Mobility management techniques for seamless connectivity • Mobile IP – MIP • Inefficient in the cases of frequent migrations • Hierarchical mobility model schemes- Hierarchical mobile IP (HMIP) • Provides faster handover for frequent migrations • Use of mobile anchor points (MAP) to group access points into different subnets *Traditional Hierarchical Networks Topology is structured like a tree Placement of MAPs are predetermined to be at the root nodes Wired and stable links CS 712

6. Problem Description • Hard to use hierarchical mobility management schemes in WMN Where to place the MAPs ? 6 and 13  6 hops away 13 and 5  4 hops away Might be good for MAP But . . . . High delay due to 5 neighbors MAPs placement is a challenging problem in WMN CS 712

7. Related Work • Not much research regarding the placement of MAPs • Fairly easy problem – place at the root node • Two different MAP placements of MAPs [13] • Difference in handover delay • Research on other issues similar to the MAP placement problem • Placing different servers at different locations for better performances [14][15] • Focus is on maximizing the network capacity • In WMN, focus  minimize the latency • Centrality type of problem [16][17][18] • Based on central indices, the centrals of the networks can be determined • First used in social network analysis • Ex- co-citation networks  most important and famous scientists CS 712

8. Mathematical Modeling (1/4) • Mathematical definition of anchor point placement • WMN as a simple undirected graph G= (V,E) • Edge delay  • Shortest path from source to destination  • Problem of selecting a set of nodes from V and a set of sub-graphs from G so that each acts as a root of to give minimum possible average delay Selection in 2 parts *Formation of sub-graphs *Selection of the root nodes CS 712

9. Mathematical Modeling (2/4) • Mathematical Approach • Handover delay calculation broken down into four parts • Movement detection delay • Router advertisement delay • Address configuration delay • Binding update delay • Local binding update delay • Global binding update delay • Total Handover Delay Local delays - not affected by the placement of MAPs Assumed to be constant  CS 712

10. Mathematical Modeling (3/4) • Analysis • Major part of the formula depends upon the average round trip time (RTT) • Can be as small as 50% for the MAP in the centre of a graph rather than at the edge • (a) shows the registration delay and the transmission delay between MN and AR • (b) depends upon the transmissi-on delay between AR and MAP • (c) is affected by the transmission delay between any MAP and any HA • Avg. delay using 6MAPs is 12.2 and using MIP is 15 -> 18% reduction • Case 1 : only one MAP in the graph • Case 2 : If each AR is a MAP • Case 3 : Selecting g MAPs from graph topology Good selection of MAP can greatly reduce the delay CS 712

11. Mathematical Modeling (4/4) • Most important factors for the placement of MAPs • Some Facts • a good MAP selection can reduce the delay upto 50% in a simple graph • HMIP with well selected MAP can reduce the handover delay as compared to MIP • Local handover possibility m is the most important factor CS 712

12. Proposed Idea • Approach • Steps: • Gather user mobility information • Based on this information, group mesh nodes into subnets • Size of the subnet > the max. one hop degree and < than max .two hop degree • Estimate the average local movement possibility m and record the number of subnets g • Calculate the combine closeness centrality value • Select the node with largest combined closeness centrality in each subnet as the MAP • Forming benchmark for the provided solution-Select the MAPs randomly • Once the subnet is determined, a mesh node can be randomly selected as the MAP • Has only possibilities to find the best combination of MAPS CS 712

13. Simulation Validation • Simulation Parameters • OMNeT ++ with INET framework • Comparisons with random selection approach • Randomly positioned 15 mesh nodes • Possibility of a node to move to a different subnet is 30% CS 712

14. Future Work & Conclusion • Future Work • Distributed scheme to dynamically orm subnets • Not always possible to group meseh nodes together based on the mobility pattern • Dynamic MAP registration scheme • Traffic thru MAPs • High delay when the # of visiting nodes becomes high • Conclusion • MAP problem  forming a tree structure in a random unplanned graph to minimize the handover delay • Hierarchical tree structures are easy to organize • Chose the important nodes as root nodes • Transform the graph into a hierarchical tree structure CS 712

15. QUESTIONS?? CS 712