Conferencing using SIP
Presenter :
Ying-Hsiang Wen Kai-Shin Chen
Member :
Hona Ou
Howard Lin
Multiparty Conferencing is an important telephony service
SIP can support quite a lot of application !!
Motivation
Outline
Architecture of SIP Conferencing
"Models for Multi Party Conferencing in SIP",J. Rosenberg, H.
Schulzrinne, Internet Draft, IETF; Nov. 2000.
Centralized Conferencing Model
"Centralized Conferencing using SIP", Kundan Singh, Gautam Nair and Henning Schulzrinne, Proceedings of IPTel, April 2001.
Distributed Conferencing Model
"Distributed Management Architecture for Multimedia Conferencing Using SIP" ,Moon-Sang Jeong, Jong-Tae Park, and Wee-Hyuk Lee, International Conference on DFMA ,2005
Comment
Why SIP
SIP vs H.323
more flexible
not proprietary
scalability
quality
popular
Any Idea??
Hello
こんにちは
你好
Bonjour
Any Idea??
Hello
こんにちは
你好
Bonjour
Any Idea??
Hello
こんにちは
你好
Audio Mixing
Mixer doesn’t send user’s sound back
send to other
silence participants
Audio Mixing
At time T, if there are M speakers in N participants
What’s the computing complexity of mixer in different situation:
participants with different or the same codec
filter in mixer or participants
Participants with different codec Filter in Mixer
send to other
silence participants
O(N) O(M)
Participants with the same codec Filter in Mixer
send to other
silence participants
O(M) O(M)
Participants with different codec Filter in participants
send to other
silence participants
O(N) O(M)
Participants with the same codec Filter in participants
send to other
silence participants
O(1) O(M)
Audio Mixing
At time T, if there are M speakers in N
participants , the computing complexity of mixer
Filter in participants Filter in Mixer
Participants with the same codec Participants with
different codec
O(M)+O(N) O(M)+O(M) O(M)+O(1) O(M)+O(N)
Full Mesh
B
A D
C
B B
B A
D
C A+C+D
For End Point
Data Processing : O(M) Bandwidth : O(M)
B
A D
C A
A+C
B C A+B
B+C
For End Point
Data Processing : O(1) Bandwidth : O(1)
A+B+C
For Server
Data Processing : O(M+N) Bandwidth : O(M+N)
Centralized
A
B
A D
C C
A+B+C
A+B B+C
For End Point
Data Processing : O(1) Bandwidth : O(1)
For Mixer Point
Data Processing : O(M+N) Bandwidth : O(M+N)
End Mixing
B
A D
C B
A D
A
B
D A+B+D
A+B+D A+B+D
A+B+D
For End Point
Data Processing : O(1) Bandwidth : O(1)
For Mixer Point
Data Processing : O(M) Bandwidth : O(M)
A+B+D
Unicast rx, Multicast tx
Large large
small small
Scaling
End point: O(1) Server : (M) End point: O(1)
Server :O(M+N) End point: O(1)
mixer : O(M+N) End point:
Bandwidth O(M)
LAN, High Bandwidth
End point:
O(M) Full Mesh
Network environment
support multicast Centralized
Server Powerful mixer
point Applied
Environment
End point: O(1) Server : (M) End point: O(1)
Server : (M+N) End point: O(1)
mixer: O(M+N) Computing
Complexity
Unicast rx Multicast tx Centralized
End Mixing Type
Attribute
Comparison with different SIP Conferencing Model
A B
Ringing INVITE
OK ACK
Conversation
BYE
OK
SIP Call Signaling Float
Signaling Procedures for Conference Initiation
(1) INVITE (2) 200 OK (3) ACK
(5) 200 OK (4) REFER
(12) 200 OK (11) INVITE
(13) ACK
(17) 200 OK (16) INVITE
(18) ACK
A B C D
Conf.
server
(6) REFER (7) 200 OK
(8) INVITE (9) 200 OK (10) ACK
(15) 200 OK (14) REFER
Performance Evaluation
Experiment in this paper
CPU: 360MHz RAM: 256MB
network: 100MB/s LAN
There are only one or two speakers in a conference
Good quality for 80 participants
Tolerable quality for 100 participants
Poor quality for 120 participants
The Bottleneck is Server CPU load!!
"Distributed Management Architecture for Multimedia Conferencing Using SIP"
,Moon-Sang Jeong, Jong-Tae Park, and Wee-Hyuk Lee, International Conference on DFMA ,2005
Distributed Conferencing Model
A conferencing framework for multimedia conferencing management should provides:
Flexibility
Integration with existing management system
High scalability
Here,high scalability is achieved by coordinated
distributed conferencing control and associated media processing.
integrates the whole conference.
manage local conferencing service
Management Mechanism for Distributed Multimedia Conferencing
Conference mixer network (CMN)
Using SIP signaling methods and its
extensions to manage a conference and CMN.
Minimum mixer participant.
New Client “C”
mixer participant NO.
max:4 min:2
Signaling procedures for invitation
(4) INVITE Focus 2 (5) 200 Hold
(6) ACK
(9) 200 OK
(8) INVITE Focus 1
(10) ACK
(7) REFER Focus 1 (1)REFER to Focus 1
(3) 200 hold
(2) INVITE Focus 1
A B
Focus 1
(ACS of A) Mixer 1
(12) REFER Mixer 2 (13) REFER Mixer 2
(24) INVIT (25) 200 OK (26) ACK
(23) 200 OK (22) BYE
(25) 200 OK
(24) INVITE Mixer 2
(26) ACK
C Focus 2 (ACS of B and C) Mixer 2
Focus 2
mixer participant NO.
max:4 min:2
Signaling procedures for leave and CMN reconfiguration
(2) 200 OK (1) BYE
B E
Focus 1
(ACS of A) Mixer 1 C
Focus 2
(ACS of B and C) Mixer 2
(3) BYE C (4) 200 OK
LEAVING C
(6) 200 OK (5) BYE
LEAVING E
(7) BYE B (8) 200 OK
(9)REFER Mixer 1
11) 200 OK
10) INVITE Mixer 1
(12) ACK
(14) 200 OK (13) BYE
(14) 200 OK (13) BYE
Performance Evaluation
Performance Evaluation
Performance Evaluation
0 50 100 150 200 250 300 350 400
10 20 30 40 50 60 70 80 90 100 110 120
Participants
Average Mixing Count
Centralized conference
Distributed conference(Full-mesh)
Average processing load of a mixer
Comment
confusing evaluations
minimum mixers participant value
PSTN
SIP
SIP
SIP
H.323 H.323 SIP/PSTN
SIP
SIP/H.323
Reference
"Models for Multi Party Conferencing in SIP",J. Rosenberg, H. Schulzrinne, Internet Draft, IETF; Nov. 2000.
"Centralized Conferencing using SIP", Kundan Singh, Gautam Nair and Henning Schulzrinne, Proceedings of IPTel, April 2001.
"Distributed Management Architecture for Multimedia Conferencing Using SIP" ,Moon-Sang Jeong, Jong-Tae Park, and Wee-Hyuk Lee, International Conference on DFMA ,2005