Upload
haley-weir
View
214
Download
0
Embed Size (px)
Citation preview
WALCOM 2013 1
Broadcasting in Conflict Aware Multi-Channel Networks
WALCOM 2013 – Feb 14, 2013
Shahin Kamali1
Joint work with Francisco Claude1, Reza Dorrigiv2, Alejandro Lopez-Ortiz1, Pawel Pralat1, Jazmin Romero1, Alejandro Salinger1, and Diego Seco3
1 David R. Cheriton School of Computer Science, University of Waterloo, Canada. 2 Faculty of Computer Science, Dalhousie University, Canada 3 Department of Mathematics, Ryerson University, Toronto, Canada4 Database Laboratory, University of A Coruna, Spain.
14/02/2013
WALCOM 2013 2
Outline
• Introduction– Broadcasting problem– Multi-channel networks• Conflict-aware model
– Problem statement• Graph families– Trees, Grids, Complete graphs
• Channel assignment
14/02/2013
WALCOM 2013 3
Broadcasting Problem
• A network is modelled by an undirected, unweighted graph
• Broadcasting problem– A single message is sent from a ‘source’ of a network to all
other vertices– Communication occurs in discrete rounds– In each round informed vertices inform ‘some’ uninformed
vertices– The goal is to find a scheme which completes in minimum
number of rounds
14/02/2013
WALCOM 2013 4
Classical Model (Telephone Model)
– In each round, each informed node can send the message to at most one neighbor
A
B
E
C
D F
A
B
D
C
E
F
14/02/2013
WALCOM 2013 5
Classical Model (Telephone Model)
• Under the telephone model– The problem is NP-hard• Remains NP-hard for planar graphs, etc. [Jakobi, et al]
• Polynomial solvable for decomposable graphs, etc. [Jakobi, et al]
– The best approximation algorithm has ratio lg n/lg lg n [Elkin, Kortsarz]
• A constant approximation?
14/02/2013
WALCOM 2013 6
t=1t=1
Multi-channel Networks
• At each round a message can be sent on a channel (multiple edges)– Frequencies in Wireless Networks
A
B
E
C
D F
A
B C
E
1 1,3
2,3
11
2
2
22,3
D F
t=21 1
22
2
221
1
14/02/2013
WALCOM 2013 7
Multi-channel Networks with Conflicts
• A conflict occurs when– Two or more neighbors of u send data to u
through the same channel in the same round– u does not receive message from that channel
t=1t=1
A
B
E
C
D F
A
B C
E
1 1,3
2,3
11
2
2
22,3
D F
t=21 1
22
2
221
1
✓✓
✗
14/02/2013
WALCOM 2013 8
Previous Work
• Geometric graphs [Mahojiran, et al, Zheng, et al]
– No theoretical analysis• An extension of telephone model– Hardness, etc.
AB
E
C
D F
1 74
35
6 810
9
14/02/2013
WALCOM 2013 9
Summary of Results
• Trees– A polynomial optimal algorithm
• Complete graph– Hardness proof
AB
E
C
D F
1 74
35
6 810
9
Single channel on each edge(simplified model)
AB
E
C
D F
1 1,32,3
11
2 22,3
2
Multiple channels on each edge(generalized model)
• Trees– Hardness proof
• Grids– A polynomial optimal algorithm
14/02/2013
WALCOM 2013 10
Simplified Model (Single Channel on Edges)
• Optimal polynomial algorithm for trees– Extension from telephone model
A
B C D E
F G H I J
K L
= 4
2 1 2 0
0 1 0 0 1
0 0
max{1+2, 2+2, 0+3}
1 2 2 3
12 3 1 3
1 1
14/02/2013
WALCOM 2013 11
Generalized Model (Multiple Channels on Edges)
• The problem is NP-hard for trees• Reduction from the set cover problem– Example:
• U = {1, 2, 3, 4, 5}• Subsets: {W = {1,2,3}, X ={2, 4}, Y = {3, 4}, Z = {4, 5}}
– There is a set cover of size k if and only if the broadcast completes in k rounds
A
1 3 52 4
W ZW,XW,Y
X,Y,Z
14/02/2013
WALCOM 2013 12
Generalized Model(multiple channels each edges)
• Polynomial algorithm for grids• Find splitters
1,2 11,3
3
2,3
23
14/02/2013
WALCOM 2013 13
Complete Graphs
• It is NP-Hard to find the optimum broadcast scheme– Even if there is only one channel on each edge– Reduction series:• Exact cover Exact cover with neighborhood
Broadcasting in complete bipartite graph Broadcasting in complete graphs
14/02/2013
WALCOM 2013 14
Hardness for Complete Graphs
• Exact Cover– Given a bipartite graph, is there a subset on left
which exactly covers all vertices on right
14/02/2013
WALCOM 2013 15
Hardness for Complete Graphs (ctd)
• Exact cover with neighborhood– Given a bipartite graph, is there a vertex u on the left an also a subset
X of vertices on the left such that all neighbors of u are exactly covered by X
– Ex: u = {a4}, X={a1,a3} is a solution
• Exact cover with neighborhood is NP-hard– Reduction from Exact cover
a1
a2
a3
a4
14/02/2013
WALCOM 2013 16
Hardness for Complete Graphs (ctd)
• The broadcasting problem is NP-hard for complete bipartite graphs • Even in the special case when
– There are a total of 2 channels– source is connected to all its neighbors with the same channel.
• Reduction from Exact Cover with Neighborhood– Broadcasting completes in two rounds iff the answer to exact coverwith neighborhood is yes
a1
a2
a3
a4
v
a2
a1
a3
a4
14/02/2013
WALCOM 2013 17
Hardness for Complete Graphs (ctd)
• The broadcasting problem is NP-hard for complete graphs under the restricted model– Reduction from broadcasting in special instances of
complete bipartite graph instances– Assuming there are at least 8 channels in the network
14/02/2013
WALCOM 2013 18
Summary of Results
• Trees– A polynomial optimal algorithm
• Complete graph– Hardness proof
AB
E
C
D F
1 74
35
6 810
9
Single channel on each edge(simplified model)
AB
E
C
D F
1 1,32,3
11
2 22,3
2
Multiple channels on each edge(generalized model)
• Trees– Hardness proof
• Grids– A polynomial optimal algorithm
14/02/2013
WALCOM 2013 19
Channel Assignment
• Assign channels to the given network– Fast communication (minimize broadcast time)– Given k channels
• Complete Graphs– Assign a single channel on all edges
• Good for broadcasting • Bad when there are more than one source
– Minimum broadcast time is 2
1
1
1
11
11
1
1
1
1
1
1
1
14/02/2013
WALCOM 2013 20
A Desired Channel Assignment
14/02/2013
WALCOM 2013 21
A Desired Channel Assignment
• k (=3) classes of vertices• Vertices in the same class
connected with the same channel
• All edges between two classes are assigned the same channel
14/02/2013
WALCOM 2013 22
A desired Channel Assignment
• Broadcasting from any node completes in 2 rounds– Having k channels and at least
k2- 2k + 1 nodes
• Gossiping completes in 3 rounds
14/02/2013
WALCOM 2013 23
Concluding Remarks
• The problem is hard, even for simple class families– Approximation algorithm
• Channel assignment– Other graph families (trees?)
14/02/2013
WALCOM 2013 24
Thanks !
14/02/2013