Minutes of H.323 Robustness Teleconference - April 12, 2000 Prepared by Maureen Stillman, Nokia

Next Teleconference: Scheduled for April 20, 2000 11AM EST USA hosted by Lucent

Action items from 4/6/2000 teleconference:

1. Post meeting minutes to mailing list 2. Jorg Ott will provide failure scenarios for gatekeepers for the Robustness Framework document. 3. Review Terry Anderson's H.225 and H.245 state machine document. We need to begin to determine what state information should be kept and how it should be restored under different failure scenarios. 4. Review IETF's SCTP and DDP protocols. Randy Smith and Qiaobing Xie of Motorola will join us on 4/20 to answer questions about these protocols.

Attending the 4/6/2000 teleconference:

Maureen Stillman maureen.stillman@nokia.com Sasha Ruditsky sasha@tlv.radvision.com Paul Jones paul.jones@ties.itu.int Terry Anderson tla@lucent.com Rex Core, Lucent (please send e-mail address) Mohash and Tim of Trillium (please send e-mail address) Timothy Sipsey Timothy.Sipsey@Aspect.com

Summary of Teleconference:

Discussion of SCTP and DDP

SCTP is about to become an RFC and can be referenced in the appropriate time frame for H.323v4, but DDP will not be ready in time to be referenced.

An advantage of SCTP versus TCP: Using TCP, a socket on one end or another can fail. SCTP knows about alternate sockets and can perform fail over.

Debate concerning Annex E and DDP:

During this debate several issues/requirements were highlighted:

1) Endpoints need to incur a minimum amount of overhead in supporting robustness protocols 2) Amount of extra network traffic generated for robustness purposes must be minimized 3) We must support robustness procedures that include conference calls 4) We need to support restoring state information "in intermediate points" that is for gatekeepers, etc.

We need to examine the overlap between DDP and Annex E. Should we use both or just one or the other? If we just use Annex E, then there is no concept of a pool of resources that can handle fail over. DDP also has a built in mechanism to store state and gives us a place to store check pointed state information. If we don't use this mechanism, then we have to invent one. We also need to reestablish state information at the intermediate points (i.e. at gatekeepers) as well in the event of a failure. DDP supports a mechanism for this.

We need to have minimum complexity at the endpoints. There is a concern about the amount of overhead incurred at the endpoint to implement DDP. Randy Stewart will supply us with some numbers on the size of a representative DDP implementation.

If we use Annex E as the transport, then we need to have end-to-end acknowledgements. We could use Annex E to guarantee delivery of the messages (hop by hop) and use DDP for resource pooling and fail over. Robustness protocols need to generate a minimum of network traffic to perform these functions.

There is also a concern that robustness for conferencing be supported. For example, if an H.245 message is sent that establishes that the floor was transferred, then if this message gets lost, then the conference call has a problem.

We are still debating the issues at this stage and need further discussion to come to some conclusion.

-- maureen

Maureen Stillman Member of Scientific Staff Voice: (607)273-0724 x62 IP Voice Networks Fax: (607)275-3610 127 W. State Street Mobile: (607)227-2933 Ithaca, NY 14850 e-mail: maureen.stillman@nokia.com www.nokia.com