Experiences with Multimedia Streaming over 2.5G and 3G Networks

Experiences with Multimedia Streaming over 2.5G and 3G

Networks

J. Chesterfield, R. Chakravorty, J. Crowcroft, P. Rodriguez, S. Banerjee

Presented by Denny Iskandar

1. Introduction

• What?– Evaluates performance of multimedia

streaming over wireless network.• Why?

– 2.5G and 3G technologies are being deployed everywhere (Europe, America, and Asia)

– Popularity of multimedia applications such as videoconferencing, Voice over IP, and audio/video broadcasting

1. Introduction

• Scope of experiment– Measurements from real networks:

• The effect of heterogeneity of network is captured by comparison across different network technologies: (GSM), GPRS, and UMTS.

• Describes the importance of cooperation between network and application using an application called vorbistreamer.

Roadmap

1. Introduction2. Network measurements3. Application measurements4. Conclusions Application layer

Presentation layer

Session layer

Transport layer

Network layer

Data link layer

Physical layer

2. Network Measurements

• Multimedia traffic characterization– Assume layered organization of media– Bandwidth used ranges between a minimum

and a maximum target rate.

Figure taken from paper.

2. Network Measurements

2.1 Propagation delay and jitter

The figure is only an approximation.

2. Network Measurements

2.1 Propagation delay and jitter

2. Network Measurements

2.1 Propagation delay and jitter– Compare with ITU recommendation for voice

communications: RTT ≤ 500 ms

GSM GPRS UMTSMean 1460 ms 220 ms 30 ms

Variance Not mentioned

80-500 ms 20-300 ms

Note: For GPRS, disabling ARQ reduces jitter at the cost of higher packet loss rate (around 3%).

Figure taken from paper.

2. Network Measurements

2.2 Capacity variation

Figure taken from paper.

2. Network Measurements

2.2 Capacity variation

2. Network Measurements

2.3 Summary– Propagation delay and delay jitter are reduced

as link capacity increases.– GPRS should disable ARQ for multimedia

application.– Sub-packet error detection [1] improves

performance.

[1] J. Chesterfield, R. Chakravorty, S. Banerjee, P. Rodriguez and I. Pratt. Transport Level Optimisations for Interactive Media Streaming Over Wide-Area Wireless Networks. In WiOpt ’04, 2004.

Figure taken from paper.

3. Application Measurements

• Use Vorbis codec– A layered codec, encodes data into a base

layer and enhancement layers

3. Application Measurements

• Design of vorbistreamer– Implements IP-based data striping; this is

used to aggregate channel bandwidth.– Uses RTP as transport protocol.– Supports interactivity constraints from fully

interactive communication to one-way streaming.

– Uses Vorbis codec.

3. Application Measurements

• Encoding techniques– For multimedia application, reliability is

disabled– Needs to add redundancy to multimedia data

to facilitate receiver-based repairs– Also involves interleaving of encoding blocks

to reduce the effect of error burst

3. Application Measurements

3.1 Intra-packet redundancy (UEP)

Header Base layer EL1 EL2 ELn...

Vorbis frame

RLC packets

FEC1 FEC2frame1 n-2 frames

FEC blocks

3. Application Measurements

• Things to note:– This “bucketing” is done at application layer,

the actual RLC packetization is done at link layer.

– Multiple layers in one packet to minimize header overhead.

3. Application Measurements

3.2 Inter-packet redundancy• Cross-packet coding

– Sends parity data in separate packets than the original data.

original data parity data

dThe greater d is:• The more effective recovery• The greater recovery delay

3. Application Measurements

3.2 Inter-packet redundancy• Cross-channel coding

– From tests, it is more likely that error occurs to A and B than to B and C

A B

C

channel 1

channel 2

3. Application Measurements

3.3 Interleaving• Intra-packet: change the order of frames in

the same packet• Inter-packet: change the order of packets

in the same channel• Inter-channel:

5channel 1

channel 2

3 17

6 4 28

Tables taken from paper.

3. Application Measurements

3.4 Comparison

4. Conclusions

• The need for cooperation between network and application.

• Encoding and organization of multimedia data is important.

• Benefits from aggregating independent channels.

• Benefits from sub-packet error detection.

5. Related works[1] J. Chesterfield, R. Chakravorty, S. Banerjee, P.

Rodriguez and I. Pratt. Transport Level Optimisations for Interactive Media Streaming Over Wide-Area Wireless Networks. In WiOpt’04: Modelling and Optimization in Mobile, Ad Hoc and Wireless Networks, 2004.

[2] R. Chakravorty, J. Chesterfield, P. Rodriguez and S. Banerjee. Measurement Approaches to Evaluate Performance Optimizations for Wide-Area Wireless Networks. In Passive and Active Network Measurement (PAM 2004) 5th International Workshop, 2004.