cing-yu chu 2012.08.06 infocom 2012. outline introduction measurement measurement results ...

35
PROFILING SKYPE VIDEO CALLS: RATE CONTROL AND VIDEO QUALITY CING-YU CHU 2012.08.06 INFOCOM 2012

Upload: anis-wilcox

Post on 01-Jan-2016

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

PROFILING SKYPE VIDEO CALLS: RATE

CONTROL AND VIDEO QUALITY

CING-YU CHU

2012.08.06

INFOCOM 2012

Page 2: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Outline

Introduction Measurement Measurement Results Modeling Skype Behaviors Analysis on TCP-friendly

Page 3: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Motivation

Skype VoIP service is well studied while video service is not

Skype video service consumes more bandwidthUp to 950 kbps

Imperative for network providers and network researchers

Page 4: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Key Questions

Q1: How does a Skype video call adapt its sending rate, video rate and quality under different network conditions?

Q2: Are Skype video calls friendly to TCP flows when they compete for network resources?

Page 5: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Methodology Measurement

Black-box approachDifferent network setting with

○ configurable packet loss,○ propagation delay○ available bandwidth

Aim to measure○ sending rate○ throughput○ RTT○ video bit rate○ frame rate

Page 6: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Methodology

ModelingRate control modelFEC modelVideo quality model

AnalysisUser back-off

○ User-level rate control schemeTCP-friendliness

Page 7: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Contribution

Measures Skype’s stationary behaviors of video callsSending rate is insensitive to packet loss

when PLR < 10%Utilization of the available bandwidth is

around 80%Overly aggressive FEC scheme, 4.5 times

the PLR

Page 8: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Contribution

Derive various models to verifyUser back-offs react fast to congestionSkype video calls are TCP-friendly

○ Due to quality-driven user back-offs

Page 9: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Measurement Test-bed

Page 10: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Experiment Design

TV news video sequence “Akiyo”From Joint Video Team (JVT)Head and shoulder movements

Virtual video camera tool Data collection

TCP-dumpSkype technical reports

Page 11: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Skype Video Call

On2 video codecVideo quantization stepVideo resolutionNumber of frames per seconds (FPS)

Page 12: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Measurements Results

Impact ofPacket LossAvailable BandwidthPropagation Delay

Page 13: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Impact of Packet Loss

PLR varies from 0% to 12% Propagation delay: 50ms 3 available bandwidth settings

250 kbps750 kbps1000 kbps

Page 14: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Impact of Packet Loss

Forward Error Correction Two states

PLR < 10% => NORM statePLR >= 10% => CONS state

Different from TCP congestion control scheme

Page 15: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Impact of Available bandwidth

Available Bandwidth varies from 50 kbps to 1000kbps

Two PLRs: 2% and 10% Propagation delay: 50ms

Page 16: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Impact of Propagation Delay Propagation delay varies from 50ms to

2000ms Available bandwidth: 500 kbps PLR: 0%

Page 17: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Modeling

Sending rate Video rate Video Quality

Page 18: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Sending Rate Model

NORM and CONS states

25 scenarios withPLR: 0% to 12%Available bandwidth: 50 kbps to 1000 kbps

γ = 0.77, μ = -10.8 and δ = 21

Page 19: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Video Rate Model

FEC ratio

FEC ratio model

Ψ = 0.15 and ω = 4.5

Page 20: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Video Quality Model

ITU-T Recommendation G.1070

frame rate

video rate

Page 21: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Video Quality Model

Frame rate model

Page 22: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Video Quality Model

a = 1.431, b = 0.02228, c = 3.759, d = 184.1, e = 1.161, h = 1.446 and g = 0.03881

Page 23: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Model Validation

Co-current UDP traffic (from iPerf)0 kbps to 600 kbps

Link capacity: 700 kbps Propagation delay: 50ms Available bandwidth for Skype

Page 24: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Model Validation

Pearson Correlation Coefficient

Sending rate Video rate Frame rate

0.9898 0.9831 0.9545

Page 25: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Analysis

Q1: How Skype video call users respond to quality degradation resulted from network impairments? How effective user back-offs are as a user-level rate control scheme?

Q2: What is the performance of a Skype video call when it competes with other Skype calls and TCP flows? Is Skype video call TCP-friendly?

Page 26: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Network Model

LTE wireless networkMultiple TCP and Skype usersM/M/1/K queue using drop-tailDownlink: 100 Mbps, uplink: 50 Mbps

Page 27: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

TCP Model

Reacts to packet loss and RTT

p = pq + pc

t = tq + tc

Page 28: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

User-level Rate Control

Video drop-off probability

Number of active users

Effective traffic generated by all users

Page 29: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

User-level Rate Control

Average traffic of each user

Expected Skype sending rate

Page 30: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Responsiveness to Loss

Page 31: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Responsiveness to Delay

Page 32: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Competition with TCP

pc = 2% and tc = 50ms

NT = # of TCP users, NS = # of Skype users

Aggregate traffic

For each Skype user

Page 33: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Competition with TCP

Scale-up factor: k

Page 34: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Conclusion

Measures Skype video trafficShows that Skype is robust against mild

packet loss and propagation delaySkype can efficiently utilize available

bandwidth Models Skype video behaviors Shows that Skype video is indeed TCP-

friendlyBased on user back-off rate control scheme

Page 35: CING-YU CHU 2012.08.06 INFOCOM 2012. Outline  Introduction  Measurement  Measurement Results  Modeling Skype Behaviors  Analysis on TCP-friendly

Q & A