Asynchronous Duty-Cycling Wireless Networks Lei Tang, Yanjun Sun, - - PowerPoint PPT Presentation

Optimizations for Route Discovery in Asynchronous Duty-Cycling Wireless Networks

Lei Tang, Yanjun Sun, Omer Gurewitz, David B. Johnson

Presentation at IEEE MASS 2012, October 2012

Asynchronous duty cycling MAC

• Asynchronous duty cycling: reduces energy

consumption and collisions.

• Receiver-initiated, e.g., RI-MAC, PW-MAC, A-MAC.

But causes poor on-demand route discovery results. active sleep

2

On-demand route discovery

1 5 7 3 4 6 2

3

An example illustrating that, in asynchronous duty- cycling network, conventional routing protocol discovers a route longer than the shortest route.

On-demand route discovery

1 5 7 3 4 6 2 {1}

4

Owing to random wakeup timings, in this example among all neighbors of source node 1, node 2 is the first neighbor waking up. Thus, Route Request is forwarded to 2.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2}

5

Similarly, Route Request is forwarded from 2 to 3 when 3 wakes up.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2} {1, 2, 3}

6

Then, Route Request is forwarded from 3 to 4.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2} {1, 2, 3} {1, 2, 3, 4}

7

Then, Route Request is forwarded from 4 to 6.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2} {1, 2, 3} {1, 2, 3, 4} {1, 2, 3, 4,6}

8

The destination receives the first Route Request.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2} {1, 2, 3} {1, 2, 3, 4} {1, 2, 3, 4,6} {1}

9

Finally, node 5 wakes up and receives a Route Request.

On-demand route discovery

1 5 7 3 4 6 2 {1} {1, 2} {1, 2, 3} {1, 2, 3, 4} {1, 2, 3, 4,6} {1} {1, 5} But node 6 will not forward the Route Request from node 5 because a node only forwards the first Route Request

• received. So the 5-hop route is discovered, rather than

the 3-hop route.

10

On-demand route discovery

1 5 7 3 4 6

How serious is the problem? Random networks: 53% longer than optimal shortest.

2 {1} {1, 2} {1, 2, 3} {1, 2, 3, 4} {1, 2, 3, 4,6} {1} {1, 5}

11

Our objective

• Improve the quality of the routes discovered.
• Our four optimization techniques:

– Fully distributed. – Work on different route metrics, such as hop-count and ETX. – Simple and no extra routing messages. – Can be used independently or in combinations.

12

Optimization 1: Delayed Selection

source destination

13

Shorter route may travel slower

Optimization 1: Delayed Selection

source destination

Wait for shorter routes;

• nly forward the shortest

14

Computing earliest Route Request forwarding time

source forwarder

…

Do not forward Route Request before: (k-1) × maxWakeupInterval – cumulativePacketBufferingTime

hop 1 hop 2 hop k

…

hop 1 hop k-1

… … …

Received a Route Request k hops from source

15

There may be a (k-1)-hop route

Optimization 2: Duty-Cycled Selection

…

Forwarder f neighbor neighbor neighbor

… …

neighbor neighbor neighbor

…

Request {s, x, y, t} Request {s, z, f} Request {s, z} Request {s, w, u, q}

• Each node wakes up once per cycle to receive.
• May receive more than one packet in one wakeup.
• Forward only the shortest Route Request when done.

Request {s, z, f} Request {s, z, f}

16

Optimization 3: Reply Updating

S Forwarder F M N Q P D wake@600 wake@620 wake@670

• riginal Route Reply:

{S, M, N, F, Q, D}

17

Optimization 3: Reply Updating

S Forwarder F M N Q P D wake@600 wake@620 wake@670

Each forwarder updates the route in Route Reply packet with the best known route.

• riginal Route Reply:

{S, M, N, F, Q, D} update Route Reply to say: {S, P , F, Q, D}

18

Optimization 4: Adaptive Backoff

Compute MAC backoffs to prioritize best route first:

• Shortest route first (for hop count h):

– e.g., backoff(h) =

• Or, e.g., best-ETX route first.

(min( ,10) % ) 10 CW r h and CW slotTime   

forwarder neighbor neighbor shorter Request {s, z} longer Request {s, w, u, q} shorter MAC backoff longer MAC backoff

19

Evaluation

• Extensive ns-2 simulations.
• Random networks with 100 nodes in (1000m × 1000m).

– Dense: each node has on average 15.6 neighbors.

• Grid network with 100 nodes.

– Sparse: each node has on average 3.6 neighbors.

• Evaluation metrics:

– Route hop-length. – Route ETX. – Route discovery latency. – Node energy consumption (i.e., duty cycle).

20

Route length in random networks

No-optimization: 53% longer Our optimizations: 0.2% longer

21

Route discovery latency in random networks

Our optimizations reduced the route discovery latency

22

Node duty cycle in random networks

Our optimizations slightly reduced route discovery node energy consumption.

23

Conclusion

• Four route discovery optimizations:

– Fully distributed. – Work on different route metrics, such as hop-count and ETX. – Simple and no extra routing messages. – Can be used independently or in combinations.

• In the evaluations, these optimizations:

– Significantly improved the routes discovered. – Reduced route discovery latency and energy consumption.

24