Beacon Vector Routing: Scalable PointtoPoint Routing in Wireless Sensornets

1
Beacon Vector Routing Scalable Point-to-Point
Routing in Wireless Sensornets
Omar Bakr ltombakr_at_cs.berkeley.edugt

• Rodrigo Fonseca, Sylvia Ratnasamy, Jerry Zhao,
  Cheng Tien Ee, David Culler,
• Scott Shenker, Ion Stoica

2
Routing in Sensor Networks

• Sensor networks pose new challenges
• Severely resource constrained
• 4K of RAM, 50 byte packets, low power, lossy
  radios
• Limitations unlikely to go away
• Newer applications will require richer
  communication among nodes
• Most current Point-to-Point algorithms
• Complex, demanding designs
• Scalability issues
• Notable exception

3
Geographic routing

• Geographic routing (e.g. GFG, GPSR, GOAFR,)
• Greedy routing using geographic coordinates
• Only local, constant state
• Two problems
• Low correlation between position and connectivity
  
• Coordinates may not be available
• BVR
• Simple, Robust, local state, no geographic
  positions

4
Beacon Vector Routing in a nutshell

• Borrow geographic routing scalability
• Greedy routing, local state
• Virtual coordinate space
• Distances in hops to a set of reference nodes
• Based on simple tree construction

5
Simple example
Coordinate Establishment
B1
B2
0
B3
6
Simple example
Route from 3,3,0 to 1,2,3
B1
B2
dist(lt3,3,0gt,lt1,2,3gt) (3-13-20-3) 6
B3
7
Simple example
Route from 3,2,1 to 1,2,3
B1
B2
1,2,3
B3
8
Evaluation

• Real Implementation
• Mica2 Testbeds

• Low level simulator
• TOSSIM

• High Level Simulator
• Algorithmic issues
• Scale
• Perfect Radios

Realism
10 100 1000 10000
Scale (nodes)
Figure from Elson et al., 2003
9
Greedy performance
routes with no flooding

• High Level simulator, 3200 nodes, random node
  placement, random beacon placement
• Density 16 neighbors/node. Load 32,000
  random-pair routes

10
Path Efficiency

• High Level simulator, 3200 nodes, random node
  placement, random beacon placement
• 10 routing beacons. Density 16/10
  neighbors/node. Load 32,000 random-pair routes

11
Performance

• Real implementation, 40 nodes, 20x50m office
  space, 5 beacons at edges
• Avg density 12, random-pair routes

12
Conclusion

• BVR is simple, scalable, robust to node failures,
  and presents efficient routes
• Using connectivity for deriving routes is good
  for low density/obstacles
• However, changes affect coordinate system
• Implementation results show it can work in real
  settings