D079 Supporting Internet Multicast Multimedia. UCL

 

Report for Final Year, 1998.17/06/98

Jon Crowcroft & Peter Kirstein

Project URL: http://www.cs.ucl.ac.uk/staff/jon/arpa/arpa.html

  1. Introduction

    2.  

    In this report, we cover the last quarter's work in the project on multicasting multimedia.

    1. Goals

 

The purpose of the proposed research project is to develop novel mechanisms for supporting and utilising Internet multicast.

 

The goals of the project are (as per the proposal):

  1. Mechanisms for supporting multicast routing protocols such as core placement strategies for CBT and PIM, Quality of Service path selection in multicast tree formation, and unicast support for reverse path calculation.

    2.  

  2. Mechanisms for utilising multicast capability in applications such as feedback control in a multicast environment, management of applications in a multicast group, and reliable multicast transport.

    3.  

  3. Mechanisms for integrating resource management with multicast forwarding, such as interface between forwarding engine and caching issues.

    4.  

  4. Securing multicast conferences.

    5.  

  5. Application Level Relays

    6.  

  6. Multimedia Recording

  1. QoS Multicast

    2.  

    The YAM work by Ken Carlberg has progressed nicely and we now have most of an implementation in gated as well as a partially complete implementation in the NS simulator. We have also collaborated with Mike Faloutsis at Toronto, who has a nice extension of the ideas in the QOSMIC work (paper at SIGCOMM 98). Also, work on random core placement, as well as integrating PGM with CBT has started in Paul White’s PhD work.

  2. Adaptation

 

We’ve pursed adaptation on three fronts based on two observations – one is that integrated services look like they are gong to take longer to deploy and be less widely available than formerly believed. The other is that pricing work and differentiated services seem to support the idea that adaptation will continue to be the best way to control quality and traffic load on the Internet. To this end, three pieces of work have been published based on implementation and simulation:

 

  1. MulTCP – this is a version of TCP that can give weighted proportional fairness to different users with no change to the intermediate routers. Two papers on this have been published – one in CCR on the analysis and simulation and the other in the HIPAPRCH workshop on measurements of the FreeBSD implementation on the CAIRN network.

  2. RLC – this is a reliable multicast transport protocol suitable for streaming data and one to may video or audio. It uses layered FEC and has a congestion control mechanism that shares the network fairly with TCP. The protocol has been implemented in C++ and in Java and has been simulated, analysed and measured on the CAIRN and on the open Internet. The work was reported in the Infocom 98 paper.

  3. Self Organising Transcoders – based on the transcoding capability in the Robust Audio Tool, and on the USG, we have built a system around a distributed algorithm and measurement of available capacity that is the bandwidth adaptation analogy for the delay adaptation In SRM.

  1. Resource Management in Forwarding Loop

We have participated in a number of experiments with QoS controlled by signalling and admission/policing/scheduling on the CAIRN Network. Observing that there are problems with RSVP for some types of applications we have started work on two new pieces of signalling technology:

  1. RSVP Proxies

  2. Dynamic Sender-based Reservation

 

We have discussed with the CAIRN community our original plans to run the UCL-CAIRN link running the CBQ and other QoS mechanisms. We would be able to carry out such experiments now - but only at the cost of interfering with the normal running of the CAIRN. The problem is that for this work we need the forwarding code ALT-Q from Sony, together with CAIRN routing code. The present versions of the CAIRN routing releases cannot support the combination. The next Release CAIRN-3, will do this; however we do not expect it to be stable until mid-July. We have negotiated with ISI that we have access to this code, and that we try to carry out the necessary experiments even before the official release. This experimentation is being co-ordinated with both ISI and NASA.

  1. Private Conferences

    2. We have completed a specification of private announcements of multicast sessions. The resulting document is being combined with the original SAP document by Mark Handley, and will move to "Experimental" status. The system permits both authentication and encryption of announcements. The privacy can be achieved both by pure symmetric encryption of the announcement, and use of Public Key systems with a Group Key approach. The initial Group Key is distributed by secure e-mail; both PGP and X.509 are supported.

     

    An implementation of the private announcements has been made in the form of an Extended SDR, called SDRv1.5 A version of this SDR will be released before the end of the month. The software has been documented, and is now being tested by a number of interested users.

  2. Application Level Relay

    3. We have updated the software for our Application Level relay, which relays between Unicast and Multicast. It now passes through SDR announcements correctly, and supports audio, video, and Whiteboard. The relay has no intelligence yet to block retransmission requests to repair SRM losses elsewhere.

     

    The relay has been packaged for easier installation. It is being used at two sites - UCL and a Dutch site.

     

  3. Multimedia Recorder

    4. We have progressed with our multimedia recorder, MMCR, which is now stable. It can be started from SDR using a plug-in which is distributed with the code. The replay can now be started from the WWW. We plan to do two further changes in the near future. First, we intend to change the control of the Recorder to use the RTSP protocols which are being standardised in the IETF. Second, we plan to change the storage file formats to a variant of ASF, rather than in the RTP format now used. This will allow the playout to be optimised to the transmission conditions at the time, rather than the conditions during recording.

     

    We have made some early studies of how to secure the server. A paper on this subject is in preparation.

  4. Publications

    5. Lorenzo Vicisano, Luigi Rizzo (Pisa) and Jon Crowcroft

    TCP-like Congestion Control for Layered Multicast Data Transfer

    ftp://cs.ucl.ac.uk/darpa/infocom98.ps

    for INFOCOM 98

     

    Philippe Oechslin, Jon Crowcroft

    Differentiated End to End Internet Services using a Weighted Proportional Fair Sharing TCP

    ftp://cs.ucl.ac.uk/darpa/multcp.ps

    in ACM CCR

     

    Isidor Kouvelas, Vicky Hardman and Jon Crowcroft

    Network Adaptive Continuous-Media Applications Through Self Organised

    Transcoding

    http://www.cs.ucl.ac.uk/staff/I.Kouvelas/publications/sot.ps.gz

    in NOSSDAV 98

     

    K Carlberg, J Crowcroft

    YAM

    In HIPPARCH 98

     

    Panos Gevros & J Crowcroft

    Multcp measurements

    In HIPPARCH 98

     

    Paul White & Jon Crowcroft

    Scheduling Multicast from Web Servers

     

    Nadia Kausar & J Crowcroft

    Reliable multicast requirements for Conference Control

    Paul White & Jon Crowcroft

    DRP – A Sender-based Dynamic Reservation Protocol

    Submitted to IEEE Network

     

  5. Staff

    6.  

    Panos Gevros

    Lambros Lambrinos

    Isidor Kouvelas

    + PhD students

    Ken Carlberg

    Paul White

    Nadia Kausar

     

  6. Future Plans

 

We have proposed a zero-cost extension to continue some of the work on the CAIRN.

See http://www.cs.ucl.ac.uk/staff/jon/arpa/97-99-extend.html