RE: [dccp] Comparison of RTP/UDP and RTP/DCCP

[louise.burness at bt.com]

One of the key questions is what actual response your adaptation module is using - how is it measureing congestion and how quickly is it responding to it. If its using TFRC it would be particulary interesting to comapre TFRC at the application and DCCP layer - how difficult is it to build the application and how well they both work etc

Assuming you have a general adpatation module (if loss over last time period T is greater than x, change to codec y)...

There are probably two aspects you would need to consider - the first is how well the different versions of the applications treat the network and other flows on it. That could probably be answered most easily through simuation - there has been plenty of work looking at how TFRC (either with with DCCP or standalone?)  behaves across a network and in the presence of other flows (eg does the congestion response really keep the network stable? does it cause competeing TCP flows to back off too much?) Try googling "ns TFRC", looking for work by Tom Phelan and Sally Floyd for example

The second question is how well the application appears to run from a user perspective, and in that case implementations may well be the best way to go. For example, a congestion response that is hand tuned to a particular application might well give a better user experiance than an geneal congestion response such as provided in DCCP (although it might be good for the application it might be dangersous for the network, hence DCCP!)

Yes I think you probably will need some kind of adaptation module to ensure you match the application output rate to the available network transmission rate, otherwise the application will be filling up a send buffer somewhere. DCCP (assuming you use with TFRC inside) should give you a reasonably smooth congestion response, and you should only need to respond to persistant congetsion at the application layer. 

Lou


-----Original Message-----
From:	dccp-bounces at ietf.org on behalf of Aylin Kantarci
Sent:	Fri 29/07/2005 09:44
To:	dccp at ietf.org
Cc:	
Subject:	[dccp] Comparison of RTP/UDP and RTP/DCCP
Hello,

 

I have implemneted a streaming software on RTP/UDP. 

How can I compare its performance with RTP/DCCP?

With RTP/UDP I am using an adaptation module for congestion control. 

I am using a scalable codec and  the adaptation module drops layers in case
of congestion.

With RTP/DCCP, do I still need an adaptation module? 

What kind of tests are required in order to successfully compare the
performance  of  UDP and DCCP?

Is simulation a better technique?  If so, is there a simulation environment
with which I can perform the tests?

Regards

 

Yrd. Doc. Dr. Aylin Kantarci

http://egeweb.ege.edu.tr/~kantarci

Bilgisayar Muhendisligi Bölümü 

Ege Üniversitesi, IZMIR

Tel: 0232 3434000-5316