broadcasting with bounded number of redundant transmissions

25
June 27, 2022 1 Broadcasting with Bounded Number of Redundant Transmissions Majid Khabbazian

Upload: tad

Post on 16-Jan-2016

24 views

Category:

Documents


1 download

DESCRIPTION

Broadcasting with Bounded Number of Redundant Transmissions. Majid Khabbazian. Outline. Assumptions Objectives Classifications The proposed algorithm Algorithm’s characteristics Conclusion. Assumptions. Single message broadcast Nodes are distributed in 2-D space - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Broadcasting with Bounded Number of Redundant Transmissions

April 21, 2023 1

Broadcasting with Bounded Number of Redundant

Transmissions

Majid Khabbazian

Page 2: Broadcasting with Bounded Number of Redundant Transmissions

2

Outline

Assumptions

Objectives

Classifications

The proposed algorithm

Algorithm’s characteristics

Conclusion

Page 3: Broadcasting with Bounded Number of Redundant Transmissions

3

Assumptions

Single message broadcast

Nodes are distributed in 2-D space

The transmission range of each node is RWe can use Unit Disk Graph (UDG) to model the network

No Synchronization

Perfect Medium Access Control (MAC)No errors or collisions

Neighbors don’t transmit at the same time

Nodes are static during the broadcast

Page 4: Broadcasting with Bounded Number of Redundant Transmissions

4

Objectives

End-to-end delay is NOT a concern

What do we care about?Full delivery

Reducing the number of transmissions

Each node has a local view of the network

Page 5: Broadcasting with Bounded Number of Redundant Transmissions

5

Flooding: A Simple Solution

FloodingEvery node transmits the first copy of received message

Pros.A simple solution

No need to have neighbor information

Requires almost no computation

Cons.All the nodes transmit the message

It can cause a large number of redundant transmissions

It can lead to significant performance degradation and network congestion

Page 6: Broadcasting with Bounded Number of Redundant Transmissions

6

A Question

Can we minimize the total number of transmissions?

This is related to fining a Minimum Connected Dominating Set (MCDS)

Finding MCDS is NP-hard even for UDGs

Good approximation algorithms?Case 1: The whole topology is known

Case 2: Each node has a local view of the networkLocal Broadcast Algorithms

Page 7: Broadcasting with Bounded Number of Redundant Transmissions

7

Local Broadcast Algorithms

ClassificationsStatic (Proactive)Dynamic (Reactive)

Static ApproachA backbone is constructed firstThe backbone is a Connected Dominating Set

Pros.Can be used for both broadcasting and unicasting

Cons.May not be good where the network topology is dynamicThe backbone is fixed in the static network

Page 8: Broadcasting with Bounded Number of Redundant Transmissions

8

Local Broadcast Algorithms (Con’d)

Dynamic ApproachThere is no backbone

Nodes decide “on-the-fly” based on their local view

Pros.The backbone changes from one network-wide broadcast to another (even for the single source)

More robust against failures than static approach

Cons.Constructed backbone may not be stable

Page 9: Broadcasting with Bounded Number of Redundant Transmissions

9

Further Assumptions

Each node has the list of its 1-hop neighborsExchanging “hello” messages

Geographical information is availableE.g., Using GPS

Relative distance may suffice

Page 10: Broadcasting with Bounded Number of Redundant Transmissions

10

Static Approach

A small size backbone can be easily constructed

Regionalizing the network

Selecting a constant number of nodes in each region

Example:Divide the network into square cells with diameter 1

At most 20 nodes have to be selected in each cell

Page 11: Broadcasting with Bounded Number of Redundant Transmissions

11

Dynamic Approach

Can we reduce the total number of transmissions in the worst case?

Is constant approximation factor achievable?

Our proposed algorithm is proven to achieve:Full delivery

Constant approximation factor

Page 12: Broadcasting with Bounded Number of Redundant Transmissions

12

Proposed Algorithm

Each node decides on its own whether or not to transmit

Before transmitting, the node removes the information attached to the message and adds the list of its 1-hop neighbors to the message

The decision is made based on a self-pruning condition called the responsibility condition

The closer, the more responsible

Page 13: Broadcasting with Bounded Number of Redundant Transmissions

13

Responsibility Condition

A node u has to transmit the message if it has a neighbor v s.t.

v has not received the message

AND

There is no node w such that w has received the message and dist(wv )< dist(uv)

Page 14: Broadcasting with Bounded Number of Redundant Transmissions

14

Example

H

G

F

A

EB

C

DA receives the message from H

A knows that E, F and G have received the message and B, C and D have not

Based on the responsibility condition A does not need to transmit the message

Page 15: Broadcasting with Bounded Number of Redundant Transmissions

15

Full Delivery

It achieves full delivery

Proof by contradiction:The broadcast will eventually terminate

Suppose there is a node that has not received the message

Consider the setS={(u,v)| u and v are neighbors, u has received the message, v has not received the message}

S is not empty

Page 16: Broadcasting with Bounded Number of Redundant Transmissions

16

Full Delivery (Con’d)

S is not emptyThere exists a pair (u’,v’) in S such that

Dist(u’,v’)<= dist(u,v)

for any pair (u,v) in S.

u’ has the highest responsibility toward v’

v’ has not receive the message

Based on the responsibility conditionu’ must have transmitted the message

Page 17: Broadcasting with Bounded Number of Redundant Transmissions

17

Approximation Factor

The proposed algorithm achieves a constant approximation factor

Sketch of proof

There are at most a constant number of transmissions in each disk with radius ¼

Transmission coverage of each node is a disk with radius 1

Each node has a constant number of neighbors that transmit the message

The number of transmission has to be within a constant factor of the optimum

Page 18: Broadcasting with Bounded Number of Redundant Transmissions

18

Approximation Factor (Con’d)

Transmitters: Blue nodes

Blue nodes are neighbors

All the nodes in the white disk will get the message after the first transmission

Blue nodes are aware of this fact

Page 19: Broadcasting with Bounded Number of Redundant Transmissions

19

Approximation Factor (Con’d)

Every blue node is responsible for a unique red node

The distance between a blue and a red node is at least ½

The number of red nods must be constant

Page 20: Broadcasting with Bounded Number of Redundant Transmissions

20

Relaxing Some of the Assumptions

Similar results can also be achieved whenNodes are distributed in 3-dimensional space

Nodes can have different transmission ranges

Nodes don’t have IDs

Geographical information is not accurateError must be less than ~0.1

Geographical information can be represented using a constant number of bits

Key Idea: Each node required to report its position to its neighbors

Page 21: Broadcasting with Bounded Number of Redundant Transmissions

21

Simulation

We compared the performance of the proposed algorithm with

Liu’s algorithm [Infocom 2006 ]

A ratio-8 approximation algorithm [Infocom 2002 ]Used as a benchmark

Page 22: Broadcasting with Bounded Number of Redundant Transmissions

22

Simulation (cont’d)

Page 23: Broadcasting with Bounded Number of Redundant Transmissions

23

Example

#nodes: 400

Trans. range: 300meter

#broadcasting nodes: 10

Page 24: Broadcasting with Bounded Number of Redundant Transmissions

24

Conclusion

Reactive broadcast algorithms are in fact powerful

Question: Can we do this without using geographical info. (or relative distances)?

The answer is YES. This can be the subject of a future talk..

Page 25: Broadcasting with Bounded Number of Redundant Transmissions

25

Thank you