GPSR: Greedy Perimeter Stateless Routing for Wireless Networks

1
GPSR Greedy Perimeter Stateless Routing for
Wireless Networks

• Brad Karp and H.T. Kung
• MobiCom 2000
• Speaker A.K. Jeng
• Date 11.Jan.05

2

• Introduction
• Algorithms
• Underlying topology
• Simulation
• Conclusion and further work

3
Introduction

• In wireless networks, communication between
  source and destination may traverses multiple
  hops.
• Distance vector and Link state algorithms require
  continual distribution of a current map of the
  entire networks topology to all routers
• When the rate change of topology or the number of
  router is high, LS and DV generates torrents of
  messages.

4

• Hierarchy
• the most widely deployed approach to scale
  routing as the number of routers increases. (ex.
  BGP)
• While wireless network lacks of well-defined AS
  boundaries.
• Caching
• reduce the overhead by sending messages on demand
  and reducing the number of hops between two ends
  (ex. DSR, ADOV, ZRP)
• Nodes must cache frequently as the rate of change
  on topology increases to update the stale state
  in nodes .

5

• The paper proposes the aggressive use geography
  to achieve scalable routing under wireless
  networks.
• measures of scalability
• Message overhead
• Delivery success rate
• Storage per node
• The following information are sufficient to make
  correct forwarding decision
• the position of packets destination
• the positions of the candidate next hops.

6

• Assumptions
• All wireless routers know their own positions
• Bidirectional radio reachability
• Wireless nodes are roughly in a plane
• Packet source can determine the locations of
  packet destinations
• A location registration and lookup service that
  maps node addresses to locations
• Queries to this systems use the same geographic
  routing system as data packets.

7
Algorithms

• Greedy Perimeter Stateless Routing algorithm
  consists of two methods
• Greedy Forwarding used if possible
• Perimeter Forwarding second choice

8
Greedy Forwarding

• All data packet are marked initially at their
  originators as greedy mode.
• Upon receiving a greedy mode packet, forward the
  packet to the neighbor closest to D.

9

• Trapped in local maximum

10
Perimeters

• When no neighbor is closer, the node marks the
  packet into perimeter mode
• x seek to route around the void using right-hand
  rule

11
When D is not reachable

• x may be adjacent to multiple faces

D
x
12

• Choice the face intersected by the line xD.
• Forward the packet around that face using the
  right-hand rule.

D
y
x

• At y, GPSR has clearly reduced the distance
  between the packet and its destination.

13
D
y
x
14
D
y
x
15
D
x
Finally, the face containing D is reached.
16
D
x
17
When D is not reachable
e0
Record e0 in the packet
D
x

• The perimeter-mode graph traversal to a reachable
  destination never sends a packet across the same
  link in the same direction twice.

18

• Return a packet to greedy mode if the distance
  from the forwarding node to D is less than that
  of Lp to D.
• Perimeter forwarding is intended to recover from
  a local maximum.

D
x
19

• The underlying topology should be a planar
• A planar graph contains no crossed edge.

D
x
20
The underlying topology

• Relative neighborhood graph, RNG

• Gabriel graph, GG

21

• RNG is a subgraph of GG
• When keeping fewer links
• Many MAC layer perform efficiently
• The contention in shared channel and hidden
  terminal problem can be reduced
• This paper uses RNG to control the underlying
  topology.

G
GG
RNG
22
Implementation issue

• Each node gathers the positions of its neighbors
  by transmitting a beacon periodically.
• Minimize the cost of beaconing
• piggyback the local sending nodes position on
  all data packets
• Reactively solicit request messages on demand.
• Upon not receiving a beacon for longer than T
  4.5B, the neighbor is seem as out-of-range or
  failed .
• The positions and set of neighbors become less
  current between beacons as the neighbors moves.
• The correct choice of beaconing interval depends
  on the rate of mobility and rage of nodes radio
  
• Replanarize the graph upon every acquisition of a
  new neighbor and every loss of a former neighbor.

23
Simulation

• Simulate GPSR in ns-2 with wireless extension.
• Motion follows the random waypoint model
• Scale of the networks

24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27

• State per node ( 200 nodes with pause time 0)
• GPSR keep 26 nodes status on average
• DSR keep 266 nodes status on average

28
Conclusion and further work

• GPSR achieve smaller per-node state, small
  routing overhead, and robust packet delivery.
• Consider the network has obstructed nodes.
• Comparison the performance when GPSR uses
  different topology control methods.
• Extend GPSR for tree-dimension space.

29
My idea

• I think the term stateless only means when the
  underlying topology control method can guarantee
  bounded node degree. i.e. RNG.
• RnG is not always good if consider shortest
  energy path

30
The END