1 / 38

Hop ID: A Virtual Coordinate-Based Routing for Sparse Mobile Ad Hoc Networks

Hop ID: A Virtual Coordinate-Based Routing for Sparse Mobile Ad Hoc Networks. BY: YAO ZHAO, YAN CHEN, BO LI, AND QIAN ZHANG PRESENTATION PREPARED BY: BABAN MAHMOOD. Organization: . Introduction Algorithm description Evaluation. Organization: . Introduction

dirk
Download Presentation

Hop ID: A Virtual Coordinate-Based Routing for Sparse Mobile Ad Hoc Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Hop ID: A Virtual Coordinate-Based Routing for Sparse Mobile Ad Hoc Networks BY: YAO ZHAO, YAN CHEN, BO LI, AND QIAN ZHANG PRESENTATION PREPARED BY: BABAN MAHMOOD

  2. Organization: • Introduction • Algorithm description • Evaluation

  3. Organization: • Introduction • Algorithm description • Evaluation

  4. Introduction • Geographic routing requires location information • It achieves excellent scalability • Does not rely on flooding • Limitations: • Dead end problem • Requires location device like GPS

  5. Organization: • Introduction • Algorithm description • Evaluation

  6. Algorithm description • Design new routing protocol to solve the dead end problem keeping efficiency even for sparse network • The algorithm utilizes a new virtual coordinates called Hop ID • A Hop ID is a multidimensional coordinates assigned based on node’s distance to landmark nodes. • Hop ID coordinates constructed base on network topology • The construction has no requirements on network density • Each node maintains a Hop ID • Each node uses a predefined hash function to hash its IP address to one of the landmarks (Location servers). • In greedy routing, a node forwards packets to a neighbor that is “nearest” to the destination in the Hop ID space.

  7. Fig 1: An example of Hop ID from the paper

  8. Distance metric • Is the key to obtain scalability • So that minimal or no flooding is necessary • Hop ID uses proposed distance function • The Hop ID distance (vector) between two nodes is calculated from the relative hop distance to the set of landmarks

  9. Hop ID description • Let Lhbe the hop distance between nodes N1 and N2 • Assume • Hop ID of N1 is (H1(1), H2(1),…, Hm(1) ) and • Hop ID of N2is (H1(2), H2(2),…, Hm(2) ) Then Max(|Hk(1) - Hk(2)|) <= Lh <= Min(|Hk(1)+ Hk(2)|) * The inequalities yield LOWER BOUND and an UPPER BOUND k k

  10. Lower bound and Upper bound • Upper bound U is not good for greedy routing • Lower bound L is good but can easily get a tie when comparing its L with the L of its neighbors. • In fig 1, B,H, and D have a tie.

  11. Power distance (Dp) • Is a modification to L to utilize more information • Dp has continuous values ,so it breaks the tie

  12. Deviation from shortest path Where: N is the paths L is the lower bound Lh is the shortest path

  13. Distance functions vs. shortest path Fig 2: Distance functions vs. shortest path. 3200 nodes used

  14. Landmark Selection • Random Landmark Selection • Peripheral Landmark Selection

  15. 1- Random Landmark Selection • Using hash function to select landmarks randomly. • For m landmarks, generate m random IDs for landmark selectioncalled ( landmark IDs) • Unique node ID can be hashed from IP address • If node’s ID is the closest one to the landmark ID, it becomes a landmark

  16. 1- Random Landmark Selection (continued) • To prevent overhead, a coordinator node C is first selected based on node stability. • Building the shortest path tree. • C generates m random landmark IDs. • Floods CENTER packet including these m IDs. • Every node adds its upstream node ID before rebroadcasting. • Aggregate landmark candidates. • The leaf node sends a CANDIDATE packet to its upstream nodes • The upstream node collects all the CANDIDATE packets and finds the best one for each landmark ID. • Iteratively, the upstream node repots to its upstream and, at last, the coordinator C will get the best candidates of the whole network.

  17. 1- Random Landmark Selection (continued) • Inform landmarks. • Now, C sends m packets instead of flooding • Each non-leaf node N knows its sub-tree, so it knows the children nodes from which each of N’s candidates are recommended. • Build Hop ID. • Each landmark floods a LANDMARK packet with its landmark ID • On receiving LANDMARK packets, each node records its hop distance to the corresponding landmark to compose its Hop ID

  18. 2- Peripheral Landmark Selection • Simple modification of Random Landmark Selection. • Detect the perimeter of the network. • C floods a CENTER packet to the entire network. • Every node finds its distance and its neighbors hop distance to the C • If the node’s hop distance is bigger than its neighbors’, it becomes a perimeter node. • So this build shortest path tree rooted at C • Similar to Random Landmark Selection (Except) • The perimeter nodes can become landmarks.

  19. Random or Peripheral Landmark Selection • Peripheral is more appealing as it requires less Landmarks • However, in high mobility, performance degrades. • Random seems to be more robust to mobility. • So, based on application: • Slow or no mobility applications like sensor net prefer Peripheral. • Otherwise, Random Selection is used.

  20. Hop ID Adjustment • In mobile environment, node may move • Hop ID has to be updated • Periodic flooding (cost and scalability) • Proposed adjustment algorithm is by applying distance vector principles • Utilizes periodic HELLO messages • Slow convergence and count to infinity problem • Split horizon with poison reverse is used. • This makes convergence faster and helps in disappearing count to infinity problem

  21. Hop ID Adjustment (Continued) • Assume N has n neighbors N1, N1, …, Nn and neighbor Ni’s Hop ID is (H1(i), H2(i), …, Hm(i)) • Then N’s new Hop ID is • Node N first calculates its new Hop ID (H1, H2, …, Hm) then broadcasts HELLO message.

  22. Hop ID Adjustment (Continued) • Each node N stores the latest Hop ID sent to the location server. • After adjustment, if the (Hop ID distance between the N’s new Hop ID and latest Hop ID > t), then N sends update packet to its location server (landmark).

  23. Landmark Management • Landmarks maintain Hop ID coordinates, so they are important nodes. • So, maintaining active landmarks is done through an algorithm: • Dealing with landmark failure or leaving. • The coordinator uses PING/PONG messages to query the landmarks • On leaving a landmark, the Coordinator uses initial landmark selection to select a new landmark. • Dealing with network partition. • Landmark HELLO messages contain timestamp • On receiving or sending HELLO messages, nodes record or include the latest timestamp of the landmark nodes. • If a node has not received a new update for a long time, then there is network partition. • New landmark can be selected as network initialized.

  24. Hop ID Greedy Routing Algorithm • Similar to geographic routing • It uses different distance metric

  25. Hop ID Greedy Routing Algorithm (continued) • Assumptions • Source knows destination’s node Hop ID and the Hop ID is included in each packet’s header • Each node know the Hop ID of its neighbors • Min(Dnd) is the minimal distance for a node R among node’s neighbors to the destination.

  26. Hop ID Greedy Routing Algorithm (continued) • Using the distance function Dp: • Source or relay (S) calculates its distance to destination (Dsd) • S calculates Min(Dnd) for R • If Min(Dnd) < Dsdthen S forwards the data packet to R. Otherwise, S is a dead end and Greedy Routing stops.

  27. The Dead End Problem

  28. Solution to The Dead End Problem • Landmark-Guided Routing Algorithm. • The Algorithm attempts to send packet toward closest landmark to the destination. • When a node E finds it’s a dead end, • It records its distance to destination (De) • Finds the nearest landmark to destination • Forwards the packet towards the landmark hop by hop letting the route entering detour mode

  29. Solution to The Dead End Problem (Continued) • The detour process continues until: • The current node is closer to the destination than E • Returning to Greedy mode • Packet reaches guiding-landmark • Landmark tries to deliver it to destination. • Packet has been forwarded more than t hops before reaching the landmark • In this case Algorithm fails

  30. Dead End Still exists !!! • The detour algorithm cannot completely resolve all dead ends • Expanding Ring flooding is used. • Floods to search some node an increases the flooding range by increasing TTL until the destination is reached.

  31. Organization: • Introduction • Algorithm description • Evaluation

  32. Evaluation • Pure Greedy Routing Algorithm is called HIR-G • Detour algorithm is called HIR-D • Expanded Ring is called HIR-E • Implemented geographic routing without location information GWL • Pure greedy geographic routing GFR

  33. Experiment Design • Most scenarios include N (N from 200 to 5100) nodes. • 2D square area (C x C) • Density λ=π x N/(C x C) is average number neighbors of node. • For 3D λ=π x N/C3 • Random Selection used. • Simulations used • NS-2 which is slow if the network size is large (e.g., 3200) • Simple scalable simulator (their simulator)

  34. Simulation results • Density • Number of nodes is 3200 • Density is from 2π to 3π

  35. Simulation results • Scalability

  36. Simulation results • Mobility

  37. Summary • Overall routing algorithm is loop-free • Hop ID system has the following steps • Voluntary node floods and builds SPF • Landmark nodes are selected randomly • Each node obtains its Hop ID • Common greedy routing algorithm

  38. Thank you

More Related