p2p iptv: hotstreaming

25
10/27/22 Copyright © 2008 1 Network and Systems Laboratory nslab.ee.ntu.edu.tw P2P IPTV: HotStreaming Polly Huang Department of Electrical Engineering National Taiwan University

Upload: rockys11

Post on 22-Jun-2015

808 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

1

Network and Systems Laboratorynslab.ee.ntu.edu.tw

P2P IPTV: HotStreaming

Polly HuangDepartment of Electrical Engineering

National Taiwan University

Page 2: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

2

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Mobile IPTV

Heterogeneous dynamics!Heterogeneous bandwidth!Heterogeneous screen size!

Page 3: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

3

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Differences

CoolStreaming HotStreamingPartnershipFormation

Pure random Preferential random(rotate)

Frame RequestScheduling

Heuristic:(fewest copies firstEarliest Deadline first)

Optimal:(min loss under bw constraint)

Codec Mpeg-4, windows WMA

Mpeg-4, interleaving MDC(bw heterogeneity)

Page 4: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

4

Network and Systems Laboratorynslab.ee.ntu.edu.tw

System Advantages

HotStreaming Advantages

PartnershipFormation

Preferential random(rotate)

Stability

Frame RequestScheduling

Optimal:(min loss under bw constraint)

Minimum loss

Codec Mpeg-4, interleaving MDC(bw heterogeneity)

Sustainable quality for heterogeneous users

Page 5: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

5

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Three ComponentsPartnership formation Multiple description codingVideo segment scheduling

Page 6: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

6

Network and Systems Laboratorynslab.ee.ntu.edu.tw

P2P Overlay Network

System Architecturepeer peer peer

peer

Partner Peers

HotStreaming Peer

Video Buffer

BufferMap

PartnershipManager

Main

Scheduler

MDCDecoder

Player

peer

Page 7: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

7

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Partnership Formation: Idea

1

4

5

2

3

67

8

9

Partnership Relationship

Every peer: 1. Periodically send2. An advertisement

message (ADV)

Upon receiving the ADV: 1. Roll a dice2. If win, select the ADV

sender to provide video segments

Consuming Peer Providing Peer

Page 8: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

8

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Qualitative Comparison SCAMP

Forwarding of ADV To all partners

Selection of partner Depending on current

#partners

TYPHOON Forwarding of ADV

To partners who have few partners

To avoid peers being isolated better stability

Selection of partner Depending on

current #partners Bounded by the max

#partners better load balance

Page 9: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

9

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Quantitative ComparisonSCAMP TYPHOON

Connected Time (sec) 59180 74320

Disconnected Time (sec) 15820 680

Disconnected Node (N) 305 18

Instability Index 51700 2580

Stability Improvement ~ 20 fold

Page 10: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

10

Network and Systems Laboratorynslab.ee.ntu.edu.tw

MDC: Idea

6420

7531

... ...

... ...Adding Redundant

Quarter -Sized Streams

6420... ...1 3 5 7

7531... ...2 4 60

O q

E f

E q

O f

MDC-STHI:MDC with Spatial-Temopral Hybrid Interpolation

Page 11: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

11

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Very Key to Mobile IPTV

Ef+Oq Of+Eq

Eq+Oq

Ef+Oq

Ef+Of+Eq+Oq

N1 N2

N3

N4 N5 N6

Ef: Full size even frameEq:1/4 size even frameOf: Full size odd frameOq: 1/4 size odd frame

Page 12: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

12

Network and Systems Laboratorynslab.ee.ntu.edu.tw

PSNR: Stefan (CIF)

15

20

25

30

35

40

45

50

1 13 25 37 49 61 73 85 97 109 121 133 145

Frame Number

PSN

R (

dB)

Full Size

STHI

Spatial

Temporal

Page 13: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

13

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Segment Request Scheduling From a peer’s partners

Available segment type (e.g. Ef, Oq) and size The peer calculates

Maximize the video quality Score of full-sized segments > quarter-sized ones

Under 2 constraints Estimated time for segment to arrive < playback time Total segments to request < estimated available

bandwidth A linear programming problem

Efficient approximation algorithm Computation < 10s milliseconds

Page 14: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

14

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Segment Request: Illustrated

1

2

3

6

2: {Ef(1), Of (1), …},

3: {Of(1), Ef (1), …}

6: {Oq(2), Eq(2), …}

LjKi

Ki

KixB

STT

Ux

ijx

L

j

ji

ji

ii

L

j

ji

ji

ji

1,1,1x0},1,0{ x

1,1x0

1,01

Subject to

Maximize

segment for typeMDC requires that schedule a :

Variables

ji

ji

L

1j

ji

11

K

1i 1

Page 15: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

15

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Testbed ExperimentsAll three components integrated

Page 16: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

16

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Impact of TYPHOON

Page 17: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

17

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Impact of MDC-STHI+Opt Schedule

Page 18: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

18

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Summary

HotStreaming Advantages

PartnershipFormation

Preferential random(rotate)

Stability

Frame RequestScheduling

Optimal:(min loss under bw constraint)

Minimum loss

Codec Mpeg-4, interleaving MDC(bw heterogeneity)

Sustainable quality for heterogeneous users

Page 19: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

19

Network and Systems Laboratorynslab.ee.ntu.edu.tw

PlanetLab ResultsSmall-scale testing on Campus Network

15-node scale testing1 server, 14 peer users1 Mbps video source for an hour

Network centric measurementScalabilityPacket loss rate (arrival rate within deadline)Control message overhead

Page 20: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

20

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Video Receiving Rate: per Peer

0

0.5

1

1.5

2

1 8 15 22 29 36 43 50 57

Time (Minute)

Mbp

s

Page 21: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

21

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Video Sending Rate: Total

0

5

10

15

20

25

1 8 15 22 29 36 43 50 57

Time (Minute)

Mbp

s Contributed by Server

Contributed by Peers

Scalability (server load)MOD ~ stream_size * #user P2P ~ stream_size * in_degree

For non-live-programs, no user limit on HotStreaming!

Page 22: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

22

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Playback Continuity

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

5 9 13 17 21 25 29 33 37 41 45 49 53 57

Time(Minute)

# L

ost /

# S

ent

CoolStreamingHotStreaming

Page 23: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

23

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Overhead

0

0.005

0.01

0.015

0.02

0.025

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57

Time(Minute)

# O

verh

ead

/ # D

ata

CoolStreaming

HotStreaming

Page 24: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

24

Network and Systems Laboratorynslab.ee.ntu.edu.tw

Questions?

Page 25: P2P IPTV: HotStreaming

04/13/23Copyright © 2008

25

Network and Systems Laboratorynslab.ee.ntu.edu.tw

ReferenceJui-Chieh Wu, Kuan-Jen Peng, Meng-Ting Lu, Chang-

Kuan Lin, Yu-Hsuan Cheng, Polly Huang, Jason Yao, Homer H. Chen, “HotStreaming: Enabling Scalable and Quality IPTV Services,” IPTV Workshop in conjunction with the 15th International World Wide Web Conference (WWW 2006), Edinburgh Scotland UK, May 2006

Meng-Ting Lu, Jui-Chieh Wu, Kuan-Jen Peng, Polly Huang, Jason J. Yao, Homer H. Chen, “Design and Evaluation of A P2P IPTV System for Heterogeneous Networks,” IEEE Transactions on Multimedia, Vol. 9, No. 8, pp. 1568-1579, Dec. 2007