Distributed Rate Control for Vid h deo Streaming with

24
Distributed Rate Control for d h Video Streaming with Intersession Network Coding Intersession Network Coding Hülya Seferoğlu, Athina Markopoulou UC Irvine

Transcript of Distributed Rate Control for Vid h deo Streaming with

Page 1: Distributed Rate Control for Vid h deo Streaming with

Distributed Rate Control for d h Video Streaming with

Intersession Network CodingIntersession Network Coding

Hülya Seferoğlu, Athina MarkopoulouUC Irvine

Page 2: Distributed Rate Control for Vid h deo Streaming with

OutlineOutlineo Introduction

o General Rate Control over NC

o Video Rate Control over NC

o Performance Evaluation

o Summary

Page 3: Distributed Rate Control for Vid h deo Streaming with

OutlineOutlineo Introduction

o Motivation o Problem Statement

G l R t C t l NCo General Rate Control over NC

o Video Rate Control over NC

o Performance Evaluation

o Summary

Page 4: Distributed Rate Control for Vid h deo Streaming with

MotivationMotivation

Network CodingVideo Streaming

Mix packets from different flows

• Transmission historyP k t d dli

Maximize video quality

Maximize Throughput

• Packet deadlines• Distortion values • Packet dependencies• Channel state

Page 5: Distributed Rate Control for Vid h deo Streaming with

MotivationMotivation

Network CodingVideo Streaming

ff Mix packets from different flows

Affect NC OpportunitiesRate

Control

Maximize Throughput

Achievable Rate Region

Time varying video content

Channel capacity

Interaction of rate control and network coding

Page 6: Distributed Rate Control for Vid h deo Streaming with

No Network Coding

AB

B

(achievable region)

A BRelayx1

x2 x1

x2

AB

x2

1

2

R/2 A1: no NC: 2x1+2x2 ≤ R

x1R/2

Page 7: Distributed Rate Control for Vid h deo Streaming with

Network Coding for Wireless (achievable region)

A BA BRelay x2

A2: NC: x1+x2+max{x1,x2} ≤ R

x1

A Bx2

B A

x1 (no NC) 1 ( NC)

R/2(R/3, R/3)max{x1, x2}

x1A,{Relay} (no NC) x1

Relay,{B} (no NC)

x1Relay,{A,B} (NC)

x1R/2x2

Relay,{A} (no NC) x2B,{Relay} (no NC)

x2Relay,{A,B} (NC)

R/2

A1: no NC: 2x1+2x2 ≤ R

Page 8: Distributed Rate Control for Vid h deo Streaming with

Network Coding for Wireless (achievable region)

A BA BRelayx1

A Bx2

B A

x1 (no NC) 1 ( NC)

max{x1, x2}

x1A,{Relay} (no NC) x1

Relay,{B} (no NC)

x1Relay,{A,B} (NC)

x2Relay,{A} (no NC) x2

B,{Relay} (no NC)

x2Relay,{A,B} (NC)

Page 9: Distributed Rate Control for Vid h deo Streaming with

Problem Statement:Rate control for video with network coding

GatewayIntermediate Node

Internet

Optimal e2e rate and network coding in the core !

Page 10: Distributed Rate Control for Vid h deo Streaming with

OutlineOutlineo Introduction

o General Rate Control over NCo Formulation

d lo Distributed Solutiono Convergence

Vid R t C t l NCo Video Rate Control over NC

o Performance Evaluation

o Summary

Page 11: Distributed Rate Control for Vid h deo Streaming with

General Rate Control over NCFormulation

A BRelayx1A B xA BRelayx1A B x2

B A

max{x1, x2}sSs

ssxxU

,)(max

τ

∑∈

U(x1) U(x2)Optimize

total utilityR1, τ1 R2, τ2

R3, τ3

x1A,{Relay} (no NC) x1

Relay,{B} (no NC)

x1Relay,{A,B} (NC)

{ 1 2}

kJiJi

kJi

IiAJiJ

sJis

Rz

xxtss

,,,

},),|({,..

τ=

= ∑∈∈ Flow

Conservation

A il blx2

Relay,{A} (no NC) x2B,{Relay} (no NC)

x2Relay,{A,B} (NC) s

JikJi

sJi

sJi

AJi Kk

kJi

zxH

qc

,..,

),(,

ξ

γτ

≤∑ ∑∈ ∈

AvailableCapacity

Interference

x1=x1A,{Relay} = x1

Relay,{B} + x1Relay,{A,B}

x =x2 = x2 + x2

max U(x1) + U(x2)Interference

Capacity constraint

x2=x2B,{Relay} = x2

Relay,{A} + x2Relay,{A,B}

Page 12: Distributed Rate Control for Vid h deo Streaming with

General Rate Control over NCDistributed solution

Rate Control Scheduling

Ssxxts

xHqxU

sJis

Ii IiAJiJ

sJi

sJi

ssJissx

s s

Ji

s

∈∀=

∑ ∑∈ ∈∈

,..

)(max},),|({

,,),(

,,η

,..

max

,

),(,,,

AAts

QR

cqc

kJi

AJi Kk

kJi

kJiJi

∈∀≤∑ ∑

∑ ∑∈ ∈

γτ

ττ

AJixx sJis

IiAJiJJiss

∈∀≥

∑∈∈

),(,0,

,

,

},),|({,

.),(,,0,

),(

AJiKkkJi

AJi Kkqc

∈∀∈∀≥

∈ ∈

τ

[ ]{ }+−+=+ sJi

kJi

sJi

sJit

skJi

skJi zxHtqtq ,,,,

,,

,, )()1( ξβParameter

Update

Page 13: Distributed Rate Control for Vid h deo Streaming with

General Rate Control over NCConvergence

C I U ( ) l ( ) C II U ( ) 4l ( )Case I: U1(x1) = log(x1), U2(x2) = log(x2)

Case II: U1(x1) =4log(x1),U2(x2) = log(x2)

Page 14: Distributed Rate Control for Vid h deo Streaming with

OutlineOutlineo Introduction

o General Rate Control over NC

o Video Rate Control over NCo Key Observationso Formulation

o Performance Evaluation

o Summary

Page 15: Distributed Rate Control for Vid h deo Streaming with

Video Rate Control over NC K b ti

o Key observation I: Video rate requirements affect the

Key observations

y qunderlying network coding opportunities.

A BRelayx1A B x2

Rate (A) Rate(B)

y

B Amax{x1, x2}

s1(A)

s2(A) s1(B)

s2(B)

Time Time

s1(A), s1(B)

Goal

s2(A) s2(B)s2(A), s2(B)

Page 16: Distributed Rate Control for Vid h deo Streaming with

Video Rate Control over NCK b i

o Key observation II: Delaying some scenes and optimizing

Key observations

y y g p gthe rate allocation create more network coding opportunities.

A Relayx1A x2 By

B Amax{x1, x2}

Rate (A) Rate(B)

s1(A)

s2(A) s1(B)

s2(B)

Time Time

s1(A), s1(B)

Goal

s2(A) s2(B)s2(A), s2(B)

Page 17: Distributed Rate Control for Vid h deo Streaming with

Video Rate Control over NCK b i

o Key observation II: Delaying some scenes and optimizing

Key observations

y y g p gthe rate allocation create more network coding opportunities.

A Relayx1A x2 By

B Amax{x1, x2}

Rate (A) Rate(B)

s1(A)

s2(A) s1(B)

s2(B)s1(A), s1(B) s2(A), s2(B)

Time Time

s1(A), s1(B)

Goal

s2(A) s2(B)s2(A), s2(B)

Page 18: Distributed Rate Control for Vid h deo Streaming with

Video Rate Control over NCFormulation

)()(..

)())((max,

fxfxts

ffxU

sJis

Ss Ffssx

s

= ∑

∑ ∑∈ ∈

δτ

Maximize total scene utilization

Scene duration)()(

,,,

},),|({,

Rz

ff

k

kJiJi

kJi

IiAJiJJiss

=

∑ ∑

∑∈∈

τ

)( ,..,

),(,

zfxH sJi

kJi

sJi

sJi

AJi Kk

kJi

qc

∑ ∑∈ ∈

ξ

γτ

Multiple scenes from a source

)()( min fxfx ss

Ff s

∈from a source

Satisfy minimum rate requirement

of each sceneof each scene

Page 19: Distributed Rate Control for Vid h deo Streaming with

OutlineOutlineo Introduction

o General Rate Control over NC

o Video Rate Control over NC

o Performance Evaluation

o Summary

Page 20: Distributed Rate Control for Vid h deo Streaming with

Performance Evaluation l Log utilities - setup

o A and B examplep

o Two sequences with six scenes

o Scenes consist of 250 packets

o Utility function of each stream; o U1=[4logx1(1), logx1(2), 4logx1(3), logx1(4), 4logx1(4), logx1(4)]1 [ g 1( ), g 1( ), g 1( ), g 1( ), g 1( ), g 1( )]o U2=[logx2(1), 4logx2(2), logx2(3), 4logx2(4), logx2(4), 4logx2(4)]

o Channel capacity of each link is considered as 10 o Channel capacity of each link is considered as 10 packets/transmission

Page 21: Distributed Rate Control for Vid h deo Streaming with

Performance EvaluationLog utilities – results

Scene Number 1 2 3 4 5 6General Rate Control

NC4.53 2.01 4.23 2.00 4.23 2.00

over NCVideo Rate Control

over NC4.98 2.04 4.66 2.00 4.66 2.00

Page 22: Distributed Rate Control for Vid h deo Streaming with

Performance Evaluationl d Real video - setup

o A and B examplepo Two sequences with six scenes o Scenes consist of QCIF size video sequences;

C h F G d M th & Carphone, Foreman, Grandma, Mother & Daughter. 30 fps, IPP.. structure, average packet size is 1000B.

o Utility function of each stream determined according to DR curve and weighted according to content; content; o w1=[0.19, 4.45, 0.18, 3.57]o w2=[2.56, 0.19, 2.56, 0.14]

Ch nn l p it f h link is nsid d s o Channel capacity of each link is considered as 1Mbps.

Page 23: Distributed Rate Control for Vid h deo Streaming with

Performance Evaluationl d lReal video - results

Page 24: Distributed Rate Control for Vid h deo Streaming with

SummarySummary

o Proposed distributed rate control schemes for o Proposed distributed rate control schemes for video streaming over wireless networks with intersession network coding.

o Improved total rate and video quality.

o Ongoing work…