Multirate Anypath Routing in Wireless Mesh Networks

1
Multirate Anypath Routing in Wireless Mesh
Networks
Rafael Laufer, Henri Dubois-Ferrière, Leonard
Kleinrock
Computer Science Department University of
California at Los Angeles
Riverbed Technology, Inc. Lausanne, Switzerland
Acknowledgments to Martin Vetterli and Deborah
Estrin
2
Loss and Instability
M. Lukac, Measuring Wireless Link Quality, 2007
3
Wireless Networks

• Different properties for the wireless medium
• Lossy and unstable links
• Limited transmission range
• Collisions and hidden terminals
• Intra- and inter-flow interference
• Broadcast nature
• Same routing paradigm for wireless networks?
• Can the broadcast medium work in our favor?

4
Anycast Forwarding

• Packet sent to multiple nodes simultaneously
• High chance of at least one node receiving it
• Node with the shortest distance forwards it on
• Coordination with overhearing and suppression

5
Anypath Routing

• Every node forwards the packet to a set of nodes
• A set of paths from the source to the destination
• This set of paths is called an anypath

6
Our Contributions

• Potential issues with single-rate anypath routing
  
• New routing paradigm for wireless networks
• Anypath routing with multiple bit rates
• Rate diversity imposes new challenges
• Introduction of a routing metric for multirate
• Routing algorithm for a single and multiple rates
• Not exponential
• Same complexity as Dijkstras and optimal
• Indoor 18-node 802.11b testbed measurements

7
Single-Rate Anypath Routing

• Under-utilization of available bandwidth
  resources
• Some hyperlinks perform well at higher rates
• Others may only work at low rates

Delivery probability
Transmission Rate
8
Single-Rate Anypath Routing

• Network disconnection at high rates
• Higher rates have a shorter transmission range
• Significant decrease in network density
• Lossier links and eventually disconnection
• Connectivity guaranteed only at low rates!

9
Multirate Anypath Routing

• Every node forwards the packet to a set of nodes
• A transmission rate for each forwarding set
• A set of paths with potentially different rates
• We call this a multirate anypath

10
Challenges

• Loss ratios usually increase with rate
• Higher rate is not always beneficial
• Shorter radio range for higher rates
• Different connectivity and density for each rate
• Higher rates
• Less spatial diversity and more hops between
  nodes
• Lower rates
• More spatial diversity and less hops between
  nodes
• How to choose both the forwarding set and rate?
• Shortest multirate anypath problem

11
Multirate Anypath Cost

• What is the cost of a multirate anypath?
• Composed of two different components
• Hyperlink cost
• Remaining cost

(r)
diJ
(r)
DJ
(r)
(r)
diJ
DJ
J
i
12
Routing Metric

• Expected transmission time (ETT)
• Average time used to transmit a packet
• Assuming a link with delivery probability
• Transmission rate and packet size
• Expected anypath transmission time (EATT)
• Tradeoff between bit rate and delivery probability

13
Remaining Cost

• Weighted average of the distances of nodes in J
• If D1?? ... ? Dn, node j is the relay with
  probability
• Weight wj(r) defined as

with
14
The Single-Rate Case

• Link-state routing protocol
• Shortest Anypath First algorithm
• Running time of O(V log V E)

.4
.5
.6
?
90
82
?
40
?
40
?
60
.7
.4
.7
.3
.8
.9
.3
.5
.5
.7
?
84
78
s
?
75
.8
?
60
.2
d
?
90
89
0
.3
.6
.6
.7
.2
.4
.2
?
87
86
85
.9
?
60
.9
?
73
15
The Single-Rate Case

• Link-state routing protocol
• Shortest Anypath First algorithm
• Running time of O(V log V E)

.2
.4
.1
?
?
?
?
.2
.1
.2
.2
.2
.4
.3
.2
.1
.2
?
s
?
.4
?
.1
d
?
0
.1
.5
.3
.3
.1
.2
.2
?
.6
?
.2
?
16
The Multirate Case

• Shortest Multirate Anypath First algorithm
• A distance estimate for each rate
• Running time of O(V log V ER)

(.5, .4)
(.4,.2)
(.6,.1)
?
73
65
66
65
?
40
30
29
24
?
40
20
?
44
84
44
43
62
43
(.4,.1)
(.7,.2)
(.7,.2)
(.8,.2)
(.3,.2)
(.9,.4)
(.3,.3)
(.5,.2)
(.5,.1)
(.7,.2)
?
70
90
70
53
65
53
s
?
(.8,.4)
58
73
58
57
64
57
?
60
38
50
38
(.2,.1)
d
?
113
70
69
70
68
(.3,.1)
(.6,.5)
0
(.6,.3)
(.7, .3)
(.4,.2)
(.2,.1)
(.2,.2)
?
57
53
56
53
(.9,.6)
?
60
30
(.9,.2)
?
43
60
43
17
Shortest Multirate Anypath First

• Why does it work?
• Three properties assuming D1?? D2 ? ... ? Dn
• Property 1
• Shortest forwarding set is of the form J 1,
  2,..., j

D1
D2
D3
18
Shortest Anypath First

• Why does it work?
• Still assuming D1?? D2 ? ... ? Dn
• Property 2
• Nodes are settled in order 1, 2,...,n
• Forwarding sets tested in order 1, 1, 2,...,
  1, 2,..., j

D1
1
1,2
1,2,3
D2
D3
19
Shortest Multirate Anypath First

• Why does it work?
• Still assuming D1?? D2 ? ... ? Dn
• Property 3
• Distance using 1 higher than distance using
  1,2, which is higher than using 1,2,3, until
  1, 2,..., j

D1
1
1,2
1,2,3
D2
Di ?? Di ? Di
Di
Di
Di
D3
20
Shortest Multirate Anypath First

• Putting it all together
• Three properties assuming D1?? D2 ? ... ? Dn
• Shortest forwarding set is of the form J 1,
  2,..., j
• Forwarding sets tested in order 1, 1, 2,...,
  1, 2,..., j
• Distance using 1 higher than distance using
  1,2, which is higher than using 1,2,3, until
  1, 2,..., j
• All properties and optimality proven in the paper

21
802.11b Indoor Testbed
22
802.11b Indoor Testbed

• Stargate microserver
• Intel 400-MHz Xscale PXA255 processor
• 64 MB of SDRAM
• Linux OS
• SMC EliteConnect SMC2532W-B PCMCIA
• IEEE 802.11b
• Prism2 chipset and HostAP driver
• Maximum transmission power of 200 mW
• Proprietary power control algorithm

23
802.11b Indoor Testbed

• Wireless mesh network
• 3-dB omni-directional rubber duck antenna
• 30-dB SA3-XX attenuator
• Weaker signal during both transmission and
  reception
• Larger distance emulated
• Network diameter
• At 11 Mbps, up to 8 hops with 3.1 hops on average
• At 1 Mbps, up to 3 hops with 1.5 hops on average

24
802.11b Indoor Testbed

• Software
• Click modular router
• MORE software package
• Modified HostAP driver
• Raw 802.11 frames
• Measure the delivery probability of each link
• 1500-byte frames
• Transmitted at 1, 2, 5.5 and 11 Mbps

25
Distribution of Delivery Probabilities
26
Evaluation Metric

• Multirate anypath routing
• Always lower cost than single-rate anypath
• Gain of multirate over single-rate anypath
• Ratio between single-rate and multirate distances
• How many times is multirate anypath better?

Di
G

Di

27
Gain of Multirate Anypath Routing
28
Transmission Rate Distribution
29
Conclusions

• Opportunistic routing paradigm for multiple rates
• Range and delivery probability change with rate
• Shortest multirate anypath problem
• Introduction of the EATT routing metric
• Shortest Multirate Anypath First algorithm
• Measurements from an indoor 802.11b testbed
• Single rate may lead to network disconnection
• Multirate outperforms 11-Mbps anypath routing by
  80 on average and up to 6.4x with full
  connectivity
• Distribution of bit rates not concentrated at any
  rate

30
Multirate Anypath Routing in Wireless Mesh
Networks
Rafael Laufer, Henri Dubois-Ferrière, Leonard
Kleinrock
Computer Science Department University of
California at Los Angeles
Riverbed Technology, Inc. Lausanne, Switzerland
Acknowledgments to Martin Vetterli and Deborah
Estrin