Interference-aware QoS Routing ( IQRouting ) for Ad-Hoc Networks
description
Transcript of Interference-aware QoS Routing ( IQRouting ) for Ad-Hoc Networks
Interference-aware QoS Routing (IQRouting) for Ad-
Hoc Networks
Rajarshi Gupta, Zhanfeng Jia, Teresa Tung, and Jean Walrand
Dept of EECS, UC Berkeley
Globecom 2005St. Louis, Missouri
EECS, UC Berkeley November 2005
2
Battalion of Tanks
Support flows with QoS Video streaming Voice calls Urgent messages
DARPA sponsored SmartNets Project
EECS, UC Berkeley November 2005
3
Interference Wired networks
Independent links Ad-hoc networks
Neighbor links interfere
Interference range > Transmission range
For simulations Tx range = 500 m Ix range = 1 km
InterferenceRange
TransmissionRange
Node A
Node D
Node C
Node B
Link 2
Link 1
EECS, UC Berkeley November 2005
4
Interference Model
Node
Link
Link
Conflict
EECS, UC Berkeley November 2005
5
Cliques Cliques
Clique= Complete Subgraph
Maximal Clique is not a subset of any other clique
Cliques in Conflict Graph Set of links that all
interfere with each other Closely related to capacity
Clique Constraints Only one link in a clique
may be active at once Flows on all links in a
clique must sum 1
A
B C
E F
D
Maximal Cliques:
ABC, BCEF, CDF
EECS, UC Berkeley November 2005
6
Available Bandwidth Available bandwidth on a link (avlbw)
Each link part of many maximal cliques Consider slack on each clique constraint Take the minimum
Available bandwidth in network/path Minimum of avlbw of all links in network/path
Key difference between wired and ad-hoc In wired, width of path determined by bottleneck
link In ad-hoc, width determined by bottleneck clique
EECS, UC Berkeley November 2005
7 Bellman’s Principle of Optimality Principle states: If optimal path from S to D
goes through A, then it follows optimal path from A to D (Bellman)
Distributed routing algorithms hinge on this principle
S AD
EECS, UC Berkeley November 2005
8
Principle of Optimality in Ad-Hoc ? Widest path from node 1
to 3 is link A (FA 1) Consider widest path
from node 1 to 5 Path A-D-E:
FA+FD+FE 1 so capacity 1/3
Path B-C-D-E: FB+FC 1, FC+FD 1,
FD+FE 1, so capacity 1/2
2
31
45
A
CB
E
Dinterference
E
CD
B
A
Connectivity Graph Conflict Graph
Does not conform with Bellman’s Principle of Optimality Hence, work with distributed heuristic algorithms
EECS, UC Berkeley November 2005
11
IQRouting at Source Link state protocol distributes available bandwidth
information Choose five candidate paths by source routing
Widest Shortest Path (WSP) WSP compliment Shortest Feasible Path (SFP) OSPF-like weighted path cost ( + used capacity) Shortest Widest Path (SWP)
Use ad-hoc versions of well-known QoS routing algorithms Account for interference among neighboring
links Clique constraints determine avlbw
EECS, UC Berkeley November 2005
12
Distributed IQRouting
Candidate paths are compared using probe packets Distributed comparison across network Nodes in path use local and current clique
information Probe rejected if lack of resources QoS metric accumulated along path
Best candidate chosen at destination
EECS, UC Berkeley November 2005
13
Comparison of Path Metric
Probe packets Evaluate clique
capacities along path Check if clique
constraints are met Accumulate path metric
(e.g. minimum of avlbw on path)
Look for bottleneck clique
FB+FC+FD+Fothers 1FD+FE+FG+Fothers 1
1
2
3
4
5
6
7
8A
C
B
D
E
H
G
EECS, UC Berkeley November 2005
15
Simulations Topology
Random 100 nodes 3 km X 3 km field Transmission 500 m Interference 1 km Flows between 5 src & 5 dest nodes
Note Random flow arrivals, durations By changing mean of flow arrival and
duration, we alter the “load” on the network
EECS, UC Berkeley November 2005
16 Comparing Admission Ratios
Competing algorithms Shortest Path OSPF ILP-based SFP Ad-Hoc SFP
2 flavors IQR-Width IQR-Cost
Results IQR performs
better
EECS, UC Berkeley November 2005
17
Grid 10X10 Grid Choose node
pairs 7 hops apart
Compare adm ratios and path length
At higher load, IQR finds longer paths with greater capacity
EECS, UC Berkeley November 2005
18
0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
2
2.5
50%
100%
0%
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2122
23
24
25
26
27
28
29
30
31
32
33
3435
36
3738
3940
41
42
43
44
45
46
47
48
49
50
X position in km
Y p
ositi
on in
km
0 kbps 1000 kbps500 kbps
Choose SourceChoose DestinationClick on bar to choose flow rateRouting…
EECS, UC Berkeley November 2005
19
0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
2
2.5
50%
100%
0%
1
2
34
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2122
23
24
25
26
27
28
29
30
31
32
33
3435
36
3738
3940
41
42
43
44
45
46
47
48
49
50
X position in km
Y p
ositi
on in
km
Choose Source, click on clear area to quit
Flow 1 from 32 to 3 at 298.9889 kbps
0 kbps 1000 kbps500 kbps
Choose Next SourceChoose DestinationClick on bar to choose flow rateRouting…
EECS, UC Berkeley November 2005
20
0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
2
2.5
50%
100%
0%
1
2
34
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2122
23
24
25
26
27
28
29
30
31
32
33
3435
36
3738
3940
41
42
43
44
45
46
47
48
49
50
X position in km
Y p
ositi
on in
km
Choose Source, click on clear area to quit
Flow 1 from 32 to 3 at 298.9889 kbpsFlow 2 from 2 to 33 at 298.9889 kbps
EECS, UC Berkeley November 2005
21
0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
2
2.5
50%
100%
0%
1
2
34
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2122
23
24
25
26
27
28
29
30
31
32
33
3435
36
3738
3940
41
42
43
44
45
46
47
48
49
50
X position in km
Y p
ositi
on in
km
Choose Source, click on clear area to quit
Flow 1 from 32 to 3 at 298.9889 kbpsFlow 2 from 2 to 33 at 298.9889 kbps
0 kbps 1000 kbps500 kbps
Choose Next SourceChoose DestinationClick on bar to choose flow rate
Flow Rejected. Insufficient Resources
EECS, UC Berkeley November 2005
22
Conclusions Multi-hop services have a long way to go
Actual capacity far lower than advertised Shortest path methods are inadequate Heuristic schemes most promising
IQRouting proposes one simple, distributed algorithm for ad-hoc networks
Performance results show significant improvement