Victor,
Apologies for the delay in getting back to you. I have been
away from my desk for some weeks. With regard to the paper, the key
aim was to investigate the delay ditribution arising from a "typical"
call over the public Internet. In particular, the delay difference
between SIP and H.323 Fast Connect was investigated. Here the key
difference is the TCP connection setup required for H.323, which
results in an extra message exchange, compared to the SIP
counterpart.
In response to your question, we considered the "dial to ring
delay", defined on the first page of the paper, to be the one
between sending an INVITE and receiving a PROVIOSIONAL RESPONSE
from a UAS. The details of the scenarios considered are in the
results sections. However, our intention was not to establish a SIP
Call Setup delay definition, but rather to examine a common
scenario. The aim of the paper was call setup delay performance
analysis, rather than to define specific SIP behaviours.
As mentioned in the earlier emails, the SIP INVITE retransmission
sequence outlined in the paper is not correct. The timeout
interval does not stop at 4 seconds, but instead continues to
double. The responsibility for this oversight is mine. However,
it does not effect the results, which comprised 95th percentile
and minimum call setup delays.
Hope this helps. Let me know if you need more.
Regards,
Tony
At 12:40 PM 3/6/01 +0000, Victor Kueh wrote:
>Dear Jonathan,
>
> Thanks for your reply. Actually I have read this paper by Tony
>and Henning several times. In that paper, call setup is defined to be the
>time between sending the INVITE and getting a provisional response. What I
>am not sure is does this provisional response have to come from a UAS or
>can it be from a stateful proxy (obviously, the call setup delay will then
>be different) - perhaps either Tony or Henning can shed some light on
>this.
>
>Thanks
>
>Victor
>
>
>> >
>> > Sorry if these questions seem trivial, but I wouild
>> > appreciate if someone out there who know the answers could help.
>> >
>> > What is the definition of call setup delay in SIP? Is it
>> > from the time the UAC sends the INVITE to the time it gets a
>> > provisional
>> > response? And does this provisional response has to be issued
>> > by a UAS or
>> > the one issued by a proxy will do (they will be different
>> > since the one
>> > issued by the proxy will reach the UAC faster)
>>
>> The spec does not define metrics as these. It would be the job of some
other
>> organization or document to define terms like "calls per second" and "call
>> setup delay" as they relate to SIP.
>>
>> I belive call setup delay generally refers to the delay between completion
>> of dial and hearing ringback. Tony Eyers and Henning Schulzrinne did some
>> work to study the distribution of these delays using SIP over real
networks;
>> see:
>>
>> http://www.fokus.gmd.de/events/iptel2000/pr/signaling_talk.pdf
>>
>> for some definitions and results.
>>
>
=================================================================
Dr Tony Eyers
Lecturer
School of Electrical, Computer and Telecommunications Engineering
University of Wollongong
Northfields Ave
Wollongong NSW 2522
Australia
Ph 61 2 42 214761
Fax 61 2 42 21 3236
Email [EMAIL PROTECTED]
==================================================================
_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors
