off by one power-save protocols presentation #4

31
Off By One Power-Save Protocols Presentation #4 Corey Andalora Keith Needels

Upload: adeola

Post on 02-Feb-2016

39 views

Category:

Documents


0 download

DESCRIPTION

Off By One Power-Save Protocols Presentation #4. Corey Andalora Keith Needels. Agenda. Final Design Graph Algorithms Test Results Future Work Conclusion. Final Design. Removed packet simulation Statistics about topology more useful - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Off By One Power-Save Protocols Presentation #4

Off By OnePower-Save ProtocolsPresentation #4

Corey Andalora

Keith Needels

Page 2: Off By One Power-Save Protocols Presentation #4

Agenda

Final Design Graph Algorithms Test Results Future Work Conclusion

Page 3: Off By One Power-Save Protocols Presentation #4

Final Design

Removed packet simulationStatistics about topology more usefulFrom statistics, latency and delivery rates can

be estimatedEasier to implement graph algorithms

Did not implement BECA / AFECAWould require a source routing algorithmGAF and Span are independent of routing

Page 4: Off By One Power-Save Protocols Presentation #4

Graph Algorithms

As nodes power cycle and move, the topology graph is updated

Run periodically through lifetime of system to collect statistics.

Page 5: Off By One Power-Save Protocols Presentation #4

Connected Pairs

B

A

Page 6: Off By One Power-Save Protocols Presentation #4

Connected Pairs

A

C

Page 7: Off By One Power-Save Protocols Presentation #4

Connected Pairs

A

C

Page 8: Off By One Power-Save Protocols Presentation #4

Connected Pairs

E

D

Page 9: Off By One Power-Save Protocols Presentation #4

Connected Pairs

D

F

Page 10: Off By One Power-Save Protocols Presentation #4

Algorithm Transition

Johnson Algorithm Finds all pairs, shortest path O( V2 log V + VE ) Can’t use it because of temporary node edges

Dijkstra Algorithm performed n2 times Finds all shortest paths to a node O( ( E + V ) log V ) with binary heap Again not taking advantage because of temporary node edges,

therefore inefficient Not necessary because hops are a weight of one.

Page 11: Off By One Power-Save Protocols Presentation #4

Algorithm Transition (cont’d)

Breadth First Search performed n2 times Determines if a path exists between two nodes O( V + E ) When all edges weigh the same, finds shortest path Perfect for our needs

Ford-Fulkerson Determines max flow of network O( VE2 ) Don’t need to temporarily add edges We use this to determine number of unique paths in system

Page 12: Off By One Power-Save Protocols Presentation #4

Test Setup

Using our AdHocSimulation program, we can run tests with a specified node type, random seed, number of nodes, radio range, maximum speed, and maximum travel distance before changing direction.

AdHocSimulation outputs a CSV file with the following columns: Time: The time (in seconds) since the start of the simulation that

this data snapshot was taken. Path probability: The percentage of pairs of nodes that currently

have a path between them. Average Hops: The average number of hops between nodes. Average number of unique paths between nodes: This is the

average maximum flow between all pairs of nodes. Percentage of nodes that have not drained all of their battery. Power remaining: The amount of power left in all nodes divided

by the total power capacity of all nodes in the system.

Page 13: Off By One Power-Save Protocols Presentation #4

Node Lifetime – 100 Node Network

0

10

20

30

40

50

60

70

80

90

100

0 50 100 150 200 250 300 350 400 450 500

Span Nodes Alive (%) Unmodified Nodes Alive (%) GAF Nodes Alive (%)

Page 14: Off By One Power-Save Protocols Presentation #4

Node Lifetime – 200 Node Network

0

10

20

30

40

50

60

70

80

90

100

0 50 100 150 200 250 300 350 400 450 500

Span Nodes Alive (%) Unmodified Nodes Alive (%) GAF Nodes Alive (%)

Page 15: Off By One Power-Save Protocols Presentation #4

Basic Test Results

Both GAF and Span offer big savings in power over no power save protocol.GAF: 190-225% increase in lifetime over

unmodified.Span: 210-280% increase in lifetime over

unmodified. Both GAF and Span have better

performance with denser networks.

Page 16: Off By One Power-Save Protocols Presentation #4

GAF – 100 vs. 400 Nodes

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600

Path Probability (%) Nodes Alive (%) Power Remaining (%) Path Probability (%) Nodes Alive (%) Power Remaining (%)

Dashed lines = results for 100 nodes, solid lines = results for 400 nodes.

Page 17: Off By One Power-Save Protocols Presentation #4

GAF – 100 Nodes

0

2

4

6

8

10

12

14

16

18

20

0 50 100 150 200 250 300 350 400 450

Average Hops # Unique Paths

Page 18: Off By One Power-Save Protocols Presentation #4

GAF – 400 Nodes

0

2

4

6

8

10

12

14

16

18

20

0 50 100 150 200 250 300 350 400 450

Average Hops # Unique Paths

Page 19: Off By One Power-Save Protocols Presentation #4

Span – 100 vs. 400 Nodes

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600

Path Probability (%) Nodes Alive (%) Power Remaining (%) Path Probability (%) Nodes Alive (%) Power Remaining (%)

Dashed lines = results for 100 nodes, solid lines = results for 400 nodes.

Page 20: Off By One Power-Save Protocols Presentation #4

Span – 100 Nodes

0

2

4

6

8

10

12

0 100 200 300 400 500 600

Average Hops # Unique Paths

Page 21: Off By One Power-Save Protocols Presentation #4

Span – 400 Nodes

0

2

4

6

8

10

12

0 100 200 300 400 500 600 700

Average Hops # Unique Paths

Page 22: Off By One Power-Save Protocols Presentation #4

Density Test Results

A higher node density improves performance in both protocols.

Quadrupling the node density gave a 70% increase in network lifetime for Span and a 25% increase in the network lifetime for GAF.

The GAF authors claim that in some cases quadrupling node density increases network lifetime by 300-600%, but we did not see this with our simulations.

Page 23: Off By One Power-Save Protocols Presentation #4

GAF – No Movement vs Movement

0

10

20

30

40

50

60

70

80

90

100

0 50 100 150 200 250 300 350 400 450 500

Path Probability (%) Nodes Alive (%) Power Remaining (%)

Path Probability (%) Nodes Alive (%) Power Remaining (%)

Dashed lines = no movement, solid lines = movement.

Page 24: Off By One Power-Save Protocols Presentation #4

GAF - Movement

0

2

4

6

8

10

12

14

0 50 100 150 200 250 300 350 400 450

Average Hops # Unique Paths

Page 25: Off By One Power-Save Protocols Presentation #4

Span – No Movement vs. Movement

0

20

40

60

80

100

120

0 50 100 150 200 250 300 350 400 450 500

Connected Pairs w/ Dead(%) Nodes Alive (%) Power Remaining (%)

Path Probability (%) Nodes Alive (%) Power Remaining (%)

Dashed lines = no movement, solid lines = movement.

Page 26: Off By One Power-Save Protocols Presentation #4

Span - Movement

0

1

2

3

4

5

6

7

8

0 50 100 150 200 250 300 350 400 450

Average Hops # Unique Paths

Page 27: Off By One Power-Save Protocols Presentation #4

Movement Test Results

Node movement seems to improve the network lifetime of both algorithms. Nodes in sparsely populated areas move into dense

areas

Path probability can vary greatly, but GAF is the clear winner here. Why? GAF has more nodes on, so you have a

greater likelihood of another active node being nearby if your grid’s active node moves out of range.

Page 28: Off By One Power-Save Protocols Presentation #4

Overall Test Results and Comments Span performs better than GAF in nearly all

measurements. While GAF can have more edge disjoint paths

between nodes, this is because it has more nodes on at any given time. We could modify Span to leave more nodes on and achieve similar results.

We were surprised by these results: It seems that GAF should an advantage with location awareness, but it appears that location awareness offers little to no advantage with power save protocols.

Page 29: Off By One Power-Save Protocols Presentation #4

Future Work

Varying transmit power simulations More efficient statistical algorithms Fusion algorithm

Run Span on active GAF nodes Investigate fairness

Page 30: Off By One Power-Save Protocols Presentation #4

Conclusion

Knowing geographical information doesn’t necessarily help you in terms of power saving.

Span seems to outperform GAF in most aspects.

A power safe protocol proves most beneficial for sensor networks, where many nodes can be dedicated to routing only.

Page 31: Off By One Power-Save Protocols Presentation #4

Questions?

Ya Xu, John Heidemann, and Deborah Estrin. “Adaptive energy-conserving routing for multi-hop ad hoc networks.” Technical Report 527, USC/Information Sciences Institute, October 2000.

Ya Xu, John Heidemann, and Deborah Estrin. “Geography-informed energy conservation for ad hoc routing,” in Proceedings of 7th Annual International Conference on Mobile Computing and Networking, pp. 70-84, July 2001.

Benjie Chen, Kyle Jamieson, Hari Balakrishnan, and Robert Morris. “Span: An energy-efficient coordination algorithm for topology maintenance in ad hoc wireless networks.” ACM Wireless Networks Journal, 8(5), 481-494, September 2002.

Stefano Basagni, Marco Conti, Silvia Giordano, and Ivan Stojmenovic. Mobile Ad Hoc Networking. John Wiley & Sons, 2004. ISBN 0-471-37313-3.