1 dude, where is my packet?. nus.soc.cs5248 ooi wei tsang 2 overview characteristics of the internet...

Post on 29-Dec-2015

215 Views

Category:

Documents

1 Downloads

Preview:

Click to see full reader

TRANSCRIPT

1

Dude, Where is My Packet?

NUS.SOC.CS5248OOI WEI TSANG

2

Overview

Characteristics of the InternetGeneral techniquesError recovery for audioEffect of loss on MPEGError recovery for MPEG

3

Loss Characteristics of The Internet

NUS.SOC.CS5248OOI WEI TSANG

4

Characteristics of Internet60-70% of paths do not show

any loss

Those with loss have an average of 4.5 – 6% packet loss

[Paxson97] End-to-end Internet packet dynamics

NUS.SOC.CS5248OOI WEI TSANG

5

Packet Loss Pattern

1000

30

NUS.SOC.CS5248OOI WEI TSANG

6

Characteristic of InternetBursts of loss are typically short

(2-3 consecutively loss packets)

Long burst do occur

Burst may occur periodically

NUS.SOC.CS5248OOI WEI TSANG

7

Wireless Link

Loss rate measured in my office 10 ~ 20%

up to 50% reported!

8

General Error Recovery Techniques

Perkins, Hodson and HardmanIEEE Network Magazine 1998

NUS.SOC.CS5248OOI WEI TSANG

9

Retransmission

1 2 3 4

1 2 4

3

3

NUS.SOC.CS5248OOI WEI TSANG

10

Redundant Data

1 2 3 4

1 2 4

2

2 3

3 4 5

53

NUS.SOC.CS5248OOI WEI TSANG

11

2 4

Error Concealment

1 2 3 4

1

12

Retransmission

for audio

NUS.SOC.CS5248OOI WEI TSANG

13

Retransmission is Bad

Need scalable feedbackNACK consumes bandwidthMay need to retransmit

frequentlyRetransmission maybe useless

NUS.SOC.CS5248OOI WEI TSANG

14

Retransmission is GoodOnly retransmit when neededExact recovery

NUS.SOC.CS5248OOI WEI TSANG

15

Scalable RetransmissionOn packet loss

T = random(0, RTT)wait for Tmulticast NACK

On receiving NACK from otherssuppress own NACK

NUS.SOC.CS5248OOI WEI TSANG

16

Retransmit when

group size is small loss rate is low large latency acceptable

17

Redundant Data

for audio

NUS.SOC.CS5248OOI WEI TSANG

18

Parity FEC

1011001

1000010

0001001

XOR 0010010

NUS.SOC.CS5248OOI WEI TSANG

19

Parity FEC

1011001

1000010

0001001

0010010

NUS.SOC.CS5248OOI WEI TSANG

20

Parity FEC Ordering

NUS.SOC.CS5248OOI WEI TSANG

21

Parity FEC Ordering

NUS.SOC.CS5248OOI WEI TSANG

22

Parity FEC Ordering

NUS.SOC.CS5248OOI WEI TSANG

23

Parity FEC Ordering

NUS.SOC.CS5248OOI WEI TSANG

24

Parity FEC Ordering

NUS.SOC.CS5248OOI WEI TSANG

25

Reed-Solomon Code

RS(n,k)

RS

NUS.SOC.CS5248OOI WEI TSANG

26

Reed-Solomon Code

NUS.SOC.CS5248OOI WEI TSANG

27

Media Specific FEC

1 2 3 4

1 2 4

2

2 3

3 4 5

53

NUS.SOC.CS5248OOI WEI TSANG

28

Interleaving

NUS.SOC.CS5248OOI WEI TSANG

29

Pro and Cons

Media Independent FECOverhead for redundant dataExact recoveryCould be computationally costly

Media Specific FECApproximate recovery only

30

Error Concealment

NUS.SOC.CS5248OOI WEI TSANG

31

Effect of Loss on AudioSpeech

Human ears can interpolateLoss up to length of phoneme can

still be tolerable

NUS.SOC.CS5248OOI WEI TSANG

32

Insertion-based RepairSpliceSilence SubstitutionNoise SubstitutionRepetition

NUS.SOC.CS5248OOI WEI TSANG

33

Other Repair Methods

InterpolationRegeneration

NUS.SOC.CS5248OOI WEI TSANG

34

Pro and Cons

Could be computationally costlyApproximate recovery onlyDoes not work well for long burst

of packet loss

35

Colin’s Recommendations

NUS.SOC.CS5248OOI WEI TSANG

36

Non-Interactive Apps

InterleavingFECRetransmission for unicast only

NUS.SOC.CS5248OOI WEI TSANG

37

Interactive ApplicationsMedia Specific FEC

NUS.SOC.CS5248OOI WEI TSANG

38

Error Concealment

Repeat

39

Effects of Packet Loss on MPEGJill Boyce and Robert GaglianelloACM Multimedia 1998

NUS.SOC.CS5248OOI WEI TSANG

40

Measurement-based StudyNeed to understand the problem

before proposing solution

NUS.SOC.CS5248OOI WEI TSANG

41

Data Gathering MethodFrom: NYC13 , Austin21, London18

To: Holmdel, NJ

NUS.SOC.CS5248OOI WEI TSANG

42

Data Gathering MethodVideo:

Two 5-mins MPEG30 fps384 kbps and 1 MbpsIBBPBBPBBPBBPBBQSIF 176x112 and SIF 352x240One row per slice

NUS.SOC.CS5248OOI WEI TSANG

43

Average Packet Loss

0

2

4

6

8

10

12

14

Packet Loss Rate

(% )

New York Texas London

1Mbps384kbps

NUS.SOC.CS5248OOI WEI TSANG

44

Packet Loss vs Time

NUS.SOC.CS5248OOI WEI TSANG

45

How to Fill Packets

B 1 frame 1 packetP 1 slice 1 packetI 1 slice 1 packet

NUS.SOC.CS5248OOI WEI TSANG

46

Packet Size Distributions

NUS.SOC.CS5248OOI WEI TSANG

47

Packet Size vs Loss Rate

NUS.SOC.CS5248OOI WEI TSANG

48

Lesson

1 slice per packet even if larger than MTU

If smaller packets means higher loss rate: fill packet until MTU

Else one slice per MTU

Alt: Change size of slice

NUS.SOC.CS5248OOI WEI TSANG

49

Frame Effected by Errors

NUS.SOC.CS5248OOI WEI TSANG

50

Correlation of Packet Loss

51

Overview of Error Recovery for Video

NUS.SOC.CS5248OOI WEI TSANG

52

Methods

RetransmissionError ConcealmentFECLimiting Error Propagation

Reference Frame Selection

NUS.SOC.CS5248OOI WEI TSANG

53

Reference Frame Selection

I P B B P

X

I

NUS.SOC.CS5248OOI WEI TSANG

54

Methods

RetransmissionError ConcealmentFECLimiting Error Propagation

Reference Frame SelectionChanging Temporal Pattern

55

Changing Temporal Patterns

Injong RheeSIGCOMM ‘98

NUS.SOC.CS5248OOI WEI TSANG

56

Basic Idea

“Better Late Than Never!”, orLate packet is still useful

NUS.SOC.CS5248OOI WEI TSANG

57

MPEG Frame Pattern

I B B P B B P

I P B B P B B

NUS.SOC.CS5248OOI WEI TSANG

58

H.261 Error Propagation

I P P P P P PX X X X X X

I P P P P P PX X X X

retransmission

loss

NUS.SOC.CS5248OOI WEI TSANG

59

H.261 Frame Pattern

I P P P P P PX

loss

NUS.SOC.CS5248OOI WEI TSANG

60

I P P P P P P

H.261 Frame Pattern

X

loss

X X

retransmission

NUS.SOC.CS5248OOI WEI TSANG

61

PTDD

Periodic Temporal Dependency Distance

Large PTDDLater DeadlinesLonger Error PropagationsLess Temporal Dependencies

NUS.SOC.CS5248OOI WEI TSANG

62

QAL

I P P P

I P P P Base Layer

EnhancementLayer

NUS.SOC.CS5248OOI WEI TSANG

63

QAL

I P P P

I P P P Base Layer + FEC

EnhancementLayer

X

NUS.SOC.CS5248OOI WEI TSANG

64

QAL + PTDD

I P P P

I P P P Base Layer

EnhancementLayer

65

SR-RTP

Feamster and BalakrishnanPV’02

NUS.SOC.CS5248OOI WEI TSANG

66

Overview

An analytical model for packet loss

SR-RTPPost Processing

NUS.SOC.CS5248OOI WEI TSANG

67

Analytic Model

p : prob. packet loss f : observed frame rate : frame drop rateSi : mean num. of packets in

frames of type iP(i) : prob. frame type is iP(F) : prob. frame is useless

NUS.SOC.CS5248OOI WEI TSANG

68

p : prob. packet loss f : observed frame rate : frame drop rateSi : mean num. of packets in frames of type iP(i) : prob. frame type is iP(F) : prob. frame is useless

NUS.SOC.CS5248OOI WEI TSANG

69

Is it accurate?

NUS.SOC.CS5248OOI WEI TSANG

70

SR-RTP

Extensions to RTP for selective reliability

NUS.SOC.CS5248OOI WEI TSANG

71

SR-RTP

0 Length

ADU Sequence Number

ADU Length

ADU Offset

Priority Layer Number

NUS.SOC.CS5248OOI WEI TSANG

72

Feamster & Balakrishnan’s ADU1 ADU = 1 Frame

NUS.SOC.CS5248OOI WEI TSANG

73

Loss Detection

LEN: 50ADU: 0ADU Size: 100ADU Off: 0

LEN: 20ADU: 0ADU Size: 100ADU Off: 80

LEN: 50ADU: 1ADU Size: 150ADU Off: 40

LEN: 50ADU: 3ADU Size: 50ADU Off: 0

NUS.SOC.CS5248OOI WEI TSANG

74

Retransmission RequestBased on priorityExample:

I-Frame: highest priorityP-Frame: higher priority if closer

to prev IB-Frame: no retransmition

NUS.SOC.CS5248OOI WEI TSANG

75

Error Concealment for VideoRepeat pixels from previous

frameInterpolate pixels from

neighbouring regionInterpolate motion vectors from

previous frame

NUS.SOC.CS5248OOI WEI TSANG

76

Inferred MV for P Frame

I P

NUS.SOC.CS5248OOI WEI TSANG

77

Inferred MV for I Frame

P P I

NUS.SOC.CS5248OOI WEI TSANG

78

Evaluation

Peak Signal to Noise Ratio (PSNR)

2552 over average square of differences in pixel values

Calculated in dB (10 log 10)

NUS.SOC.CS5248OOI WEI TSANG

79

Performance of SR-RTP

80

Today’s Summary

NUS.SOC.CS5248OOI WEI TSANG

81

How to recover packet lossRetransmissionFECError Concealment

NUS.SOC.CS5248OOI WEI TSANG

82

How to limit the damage of error InterleavingKey frame selectionChange reference frame

top related