22
Peter Parnes, CDT 1/2 Media Scaling of IP-Multicast Streams in Heterogeneous Networks Peter Parnes LTU-CDT/Marratech Roxy Workshop 980921-23

Peter Parnes, CDT1/22 Media Scaling of IP-Multicast Streams in Heterogeneous Networks Peter Parnes LTU-CDT/Marratech Roxy Workshop 980921-23 Media Scaling

  • View
    216

  • Download
    0

Embed Size (px)

Citation preview

Peter Parnes, CDT 1/22

Media Scaling of

IP-Multicast Streams in Heterogeneous Networks

Peter Parnes

LTU-CDT/Marratech

Roxy Workshop 980921-23

Media Scaling of

IP-Multicast Streams in Heterogeneous Networks

Peter Parnes

LTU-CDT/Marratech

Roxy Workshop 980921-23

Peter Parnes, CDT 2/22

Overview Background Current Problem Proposed Solutions mStar Current Status in the Internet Summary

Background Current Problem Proposed Solutions mStar Current Status in the Internet Summary

Peter Parnes, CDT 3/22

Background

“Broadcasts” of real-time media on the Internet is becoming more and more important.

It is very central to the Roxy project. If the used system shall scale, IP-

multicast HAS to be used!

“Broadcasts” of real-time media on the Internet is becoming more and more important.

It is very central to the Roxy project. If the used system shall scale, IP-

multicast HAS to be used!

Peter Parnes, CDT 4/22

Requirements and Restrictions

Best-effort delivery Reliability not required Applications have to be adaptive, i.e.

have to adapt to network congestion and be able to handle different configurations.

Best-effort delivery Reliability not required Applications have to be adaptive, i.e.

have to adapt to network congestion and be able to handle different configurations.

Peter Parnes, CDT 5/22

Which bandwidth should be used when transmitting a real-time media stream over heterogeneous networks?

Which bandwidth should be used when transmitting a real-time media stream over heterogeneous networks?

Internet

Sender

100Mbps

LocalReceiver

100Mbps

InternetReceiver

500Kbps

ISDNReceiver

128Kbps

Peter Parnes, CDT 6/22

Proposed solutions

Max/Min client bandwidth Simulcast Network transcoders Receiver driven Layered Multicast - RLM Bandwidth Guessing - TCP friendly Active Networks Active Services

Max/Min client bandwidth Simulcast Network transcoders Receiver driven Layered Multicast - RLM Bandwidth Guessing - TCP friendly Active Networks Active Services

Peter Parnes, CDT 7/22

Max/Min client bandwidth

Just ignore some set of receivers Send the stream with high bandwidth

Ignore low bandwidth receivers Send the stream with low bandwidth

Force high bandwidth receivers to use low quality

Does not take congestion into account

Just ignore some set of receivers Send the stream with high bandwidth

Ignore low bandwidth receivers Send the stream with low bandwidth

Force high bandwidth receivers to use low quality

Does not take congestion into account

Peter Parnes, CDT 8/22

Simulcast Send the same stream with different

encodings from the sender and let the receivers choose what they want to receive. Can be very expensive CPU wise Wastes bandwidth on shared links

Does not take congestion into account in the way it is being used today.

Used in mStar (more later)

Send the same stream with different encodings from the sender and let the receivers choose what they want to receive. Can be very expensive CPU wise Wastes bandwidth on shared links

Does not take congestion into account in the way it is being used today.

Used in mStar (more later)

Peter Parnes, CDT 9/22

Network Transcoders A common approach is to deploy

transcoders on the boundaries between different networks. Transcoding, mixing, downscaling

E.g. transcode MJPEG to H.261 when the traffic leaves a campus (high bandwidth network).

E.g. mStar mTunnel can transcode when tunnelling mcast data (more later).

A common approach is to deploy transcoders on the boundaries between different networks. Transcoding, mixing, downscaling

E.g. transcode MJPEG to H.261 when the traffic leaves a campus (high bandwidth network).

E.g. mStar mTunnel can transcode when tunnelling mcast data (more later).

Peter Parnes, CDT 10/22

Receiver driven Layered Multicast - RLM

Divide the stream into a hierarchy of exclusive additive layers

Each layer is multicasted to a different group

loop:

if no_congestion then

join next group to get higher layer

else

leave group to drop highest layer

Divide the stream into a hierarchy of exclusive additive layers

Each layer is multicasted to a different group

loop:

if no_congestion then

join next group to get higher layer

else

leave group to drop highest layer

Peter Parnes, CDT 11/22

RLM Problems How to detect congestion caused by my

tests or by others Shared learning proposed Does it scale?

Today long timeout in mcast forwarding trees Might lead to false interpretation of the

current situation Is not “nice” to TCP

How to detect congestion caused by my tests or by others Shared learning proposed Does it scale?

Today long timeout in mcast forwarding trees Might lead to false interpretation of the

current situation Is not “nice” to TCP

Peter Parnes, CDT 12/22

Bandwidth Guessing In early 97 a proposal called “TCP-

Friendly” was distributed. Describes a way of estimating the

bandwidth between a sender and a receiver based on RTT and current packet drop.

Takes TCP into account and will be a “nice” participant in the network

In early 97 a proposal called “TCP-Friendly” was distributed.

Describes a way of estimating the bandwidth between a sender and a receiver based on RTT and current packet drop.

Takes TCP into account and will be a “nice” participant in the network

Peter Parnes, CDT 13/22

BW Guessing Problems

Hard to calculate RTT accurately Works only for “broadcast” situations Not very tested yet

Hard to calculate RTT accurately Works only for “broadcast” situations Not very tested yet

Peter Parnes, CDT 14/22

Active Networks

The latest “buzzword” network research topic (since ATM is practically dead)

Basic idea: Allow injection of small programs into

network nodes Network nodes perform computations on

user data

The latest “buzzword” network research topic (since ATM is practically dead)

Basic idea: Allow injection of small programs into

network nodes Network nodes perform computations on

user data

Peter Parnes, CDT 15/22

Active Networks...

Two Different Approaches Code and control is handled out-of-band Each packet carries miniature programs

(capsules) Allows networks to be modified “on-

demand” Opens a completely new area for real-

time media scaling

Two Different Approaches Code and control is handled out-of-band Each packet carries miniature programs

(capsules) Allows networks to be modified “on-

demand” Opens a completely new area for real-

time media scaling

Peter Parnes, CDT 16/22

Active Networks...

Issues: Safety, fairness, appropriate architecture,

common programming model, robustness Status:

At the very beginning A very “political” problem

“I dare You to run code in my router!”

Steve Deering - Cisco (currently :-)

Issues: Safety, fairness, appropriate architecture,

common programming model, robustness Status:

At the very beginning A very “political” problem

“I dare You to run code in my router!”

Steve Deering - Cisco (currently :-)

Peter Parnes, CDT 17/22

Active Services

Deploy user controllable programs-pads in the network.

Users can deploy their own transcoding programs and can easily up-grade these when needed

A system for this is currently being deployed and tested on Berkeley Campus

Deploy user controllable programs-pads in the network.

Users can deploy their own transcoding programs and can easily up-grade these when needed

A system for this is currently being deployed and tested on Berkeley Campus

Peter Parnes, CDT 18/22

mStar mTunnel contains support for media-

aware rescaling, transcoding, mixing and switching of audio and video. Could be easily be extended for general

transcoding between different mcast groups. Simulcast is currently being used in mStar

Pro for the electronic corridor. BW Guessing and RLM based

approaches should be further investigated.

mTunnel contains support for media-aware rescaling, transcoding, mixing and switching of audio and video. Could be easily be extended for general

transcoding between different mcast groups. Simulcast is currently being used in mStar

Pro for the electronic corridor. BW Guessing and RLM based

approaches should be further investigated.

Peter Parnes, CDT 19/22

Status in Internet

Almost all traffic is still sent using Unicast - transcoding at the server

Network transcoders probably most common

Simulcast less common than one could imagine (lack of good support in today’s applications).

Almost all traffic is still sent using Unicast - transcoding at the server

Network transcoders probably most common

Simulcast less common than one could imagine (lack of good support in today’s applications).

Peter Parnes, CDT 20/22

Summary A number of more or less proposed

solutions: Max/Min client bandwidth Simulcast Network transcoders Receiver driven Layered Multicast - RLM Bandwidth Guessing - TCP friendly Active Networks Active Services

Still a lot of research needed

A number of more or less proposed solutions: Max/Min client bandwidth Simulcast Network transcoders Receiver driven Layered Multicast - RLM Bandwidth Guessing - TCP friendly Active Networks Active Services

Still a lot of research needed

Peter Parnes, CDT 21/22

Questions?

[email protected]

http://www.cdt.luth.se/~peppar/

http://www.cdt.luth.se/mStar/

http://www.marratech.com/

[email protected]

http://www.cdt.luth.se/~peppar/

http://www.cdt.luth.se/mStar/

http://www.marratech.com/

Peter Parnes, CDT 22/22

Multicast vs. Unicast