Dual-Region Location Management for Mobile Ad Hoc Networks Yinan Li, Ing-ray Chen, Ding-chau Wang Presented by Youyou Cao
Introduction MANET: A self-organizing and self-configuring infrastructureless network of mobile devices connected by Wireless Problem: No good scalable location management for MANET yet!
Outline Introduction System Design Performance Model Evaluation Conclusion
Existing location management scheme Each node is assigned a home region(serve as location servers) through hashing Drawback: need to contact location server even source and destination nodes are close to each other Home region of destination node Destination node Source node node
How to improve? Periodically exchange location info with neighboring nodes in its local region If a node in the local region knows the location of destination node Local region of source node Source node Destination node Local region of destination node - Only local location info of the neighbor is needed
How to improve? Even better, source node might be within the local region of the destination node Source node Destination node Local region of destination node - Only need local info from the node itself!
How to improve? Further question: how to define home and local region size? Previous work: - Define home region size statically at design time Our work: - Dynamically determine the optimal home region size and local region size - in order to minimize the overall network cost incurred by location management and data packet delivery
System Design Assumptions: - -Mobile nodes knows their location, moving direction, moving speed via GPS - Density of mobile nodes are high enough so there is at least one location server in each node’s home region
Dual-Region Mobility Management (DrMoM) Global Partition: Equally sized rectangular region Each node is permanently assigned a home region by hashing Home region center is fixed, Home region size is dynamically chosen based on node mobility and service characteristics.
Dual-Region Mobility Management Local region center moves when node moves, local region size chosen dynamically Local region location updates follow a threshold-based approach, ie, notice neighbors when its current location is outside of the transmission rage of its last updated location
Key Parameter: home region size
Key Parameters: local region size
Location Table
Greedy geographical packet forwarding For each hop, select the node that is closest to the destination within its one hop neighbors
Performance Model Goal: minimize total communication cost incurred by DrMoM - Total number of wireless transmissions per time unit Impact of using this metric: - Small saving in cost can be significant over time - -Larger probability of successful packet deliveries - -Shorter average packet delay - -Maximize the lifetime of a MANET
Performance Model Assumptions: - Use modified random way point mobility model to simulate the movement of mobile nodes - Hash function maps any mobile node uniformly to any rectangular region with equal probability
Notation Table
Notation Broadcast cost b(R): the number of wireless transmissions to cover the entire region
Location Update Cost
Location Query Cost the cost for local region location query: broadcast and collect reply from local neighbors the cost for home region location query: send request and get response from home region Probability that local region location query fails Node density Probability that a neighbor of S is also in local region or home region of D
Data Packet Delivery Cost Upper bound: p1(p2) : – probability that S is within the local(home) region of D
Home Region Maintenance Cost When node B enter the home region of node A, A’s home region nodes broadcast the message to their neighbors
Total Communication Cost Data Packet RateLocal region location update rate Home region location update rate Home region maintenance rate
Performance Evaluation
Performance Comparison Total communication cost incurred per time unit by DrMoM vs SLURP and RUDLS as a function of φ, ν, n
Conclusion
Future work Consider other mobility model rather than random movement Use stochastic Petri Net modeling technique How to select trustworthy nodes as location server when there is attack Extend the design notion to other location- based services in MANET
Thank you! Questions?