This document describes a Round Trip Delay Control (RTDC) algorithm to implicitly control SIP overload without requiring SIP header modifications. The algorithm models the interaction between an overloaded SIP server and its upstream server as a feedback control system. A PI controller regulates the retransmission rate to maintain the round trip delay below a target value, preventing overload propagation. OPNET simulations validate that the RTDC algorithm can effectively cancel overload under two scenarios, outperforming other algorithms. The control-theoretic approach allows any carrier to freely implement implicit SIP overload control in their servers.

1. IEEE ICC 2011
Yang Hong, Changcheng Huang, and James Yan
Department of Systems and Computer Engineering,
Carleton University, Ottawa, Canada

2. Wh t i SIP?
What is SIP?
 Session Initiation Protocol
 protocol that establishes,
Internet
manages (multimedia)
sessions [RFC 3261]
 used for VoIP presence &
VoIP,
video conference
Proxy
Server
Proxy
Server
 SIP consists of two basic
elements
l
t
 UA (User Agent) and P-Server
(Proxy Server)
 About 1000 companies produce
SIP products
 Microsoft’s Windows
Messenger (≥4 7) i l d SIP
M
(≥4.7) includes
UA
UA
Simplified SIP Network Configuration
2

3. IMS SIP Server Overload – A
f
h ll
Performance Management Challenge
 3GPP has adopted SIP
 as the basis of IMS architecture
 Problem: Server(s) cannot complete
the processing of requests under
overload conditions
 Multiple causes: Insufficient
p
capacity, Component Failures,
Unexpected traffic surges, DOS
attacks [RFC 5390]
 Impact: Performance degradation,
drop in throughput, revenue loss,
network collapse
Simplified
Si lifi d IMS C t l L
Control Layer O
Overview
i
3

4. Why Worry About SIP Message Retransmission?
Why Worry About SIP Message Retransmission?
 Retransmission built-in to maintain SIP reliability
y
against message loss
 Loss is detected as long delay in acknowledgment
 Surge in user demand can cause SIP server
overload and long delay to acknowledge SIP
messages
 Long delays may trigger more retransmissions and a
positive feedback exacerbating server overload
4

5. C t ib ti
f Thi P
Contributions of This Paper
 Using control-theoretic approach to
g
pp
 model the interaction of overloaded server and its
upstream server as a feedback control system
 Proposing Round Trip Delay Control (RTDC) algorithm
(a PI rate control algorithm) to mitigate the overload by
 regulating retransmissions
 clamping round trip delay below a desirable target value
 Performing OPNET simulations under two typical
overload scenarios to
 validate RTDC (implicit SIP overload control) algorithm
5

6. Outline
 SIP Retransmission Mechanism Overview
 Related Work on SIP Overload Control
 Queuing Dynamics of Overloaded Server
 Control-Theoretic Design for Overload Control Based
on R
Round-Trip Delay
dTi D l
 Performance Evaluation to Validate RTDC SIP
Overload Control Algorithm
 Conclusions
6

7. Typical SIP Procedure
7

8. Retransmission Mechanism
Retransmission Mechanism
 Purpose: Confirmation of successful transmission
P-servers
between UA and UA via P servers
 Two Types:
 Hop by Hop
First retransmission after T1 , subsequent one is 2
times previous interval. Total intervals up to 64 x T1
(maximum 6 retransmissions). Default T1 = 0.5 s.
 End-to End
First t
Fi t retransmission after T1 , subsequent one is 2 ti
i i
ft
b
t
i
times
previous interval up to a maximum of T2 . Total
intervals up to 64 x T1 (maximum 11 retransmissions).
Default T2 = 4.0 s.
8

9. Related Work on Overload Control
 Most of existing overload control solutions adopt
push-back mechanism
 cancel the overload effectively

by introducing overhead to advertise upstream servers to
 reduce message sending rate
d
di
t
 produce overload propagation from sever to server
until end-users
 block a large amount of calls unnecessarily

cause revenue loss of service providers
• Our Proposal: Reduce retransmission rate only to
mitigate overload
 by maintaining original message rate to

keep the revenue of service providers
9

10. SIP Overload Control Mechanism Classification
Figure 3. The classification for the existing SIP overload control schemes
Y. Hong, C. Huang, and J. Yan, “A Comparative Study of SIP Overload Control Algorithms,”
Network and T ffi E i
N
k d Traffic Engineering i E
i in Emerging Di ib d C
i Distributed Computing A li i
i Applications, Edi d b
Edited by
J. Abawajy, M. Pathan, M. Rahman, A.K. Pathan, and M.M. Deris, IGI Global, 2012, pp. 1‐20.
10

11. Queuing Dynamics of Overloaded Server
Q
g y
100Trying response
Invite request 1(t)
r1(t)
r2' (t )
Timer fires
Message buffer

q1(t)
Reset timer
Timer expires

qr1(t)
Invite request
Server 2
2(t)
2
1

r2(t)
Server 1
2(t)
q2(t)
1(t)
100Trying response
Timer starts
Ti
Timer buffer
Queuing dynamics of Server 2
Queuing dynamics of Server 1

q 2 (t )  2 (t )  r2 (t )   2 (t )   2 (t )

q1 (t )  1 (t )  r1 (t )  r2' (t )  1 (t )  1 (t )
(1)
(2)
Notation: 1(t) original message rate, r1 (t) message retransmission rate,
2(t) service rate 1 (t) response rate q1 (t) queue size
rate,
rate,
Overload Scenario: Server slowdown at Server 2 due to routine maintenance
Overload Collapse: 2(t)   2(t) > 2(t)  q 2 (t )  0 (see Eq. (1))  q2(t) 

ti
trigger r''2(t)  r2(t) i
increases q2(t) more quickly
i kl

Overload Propagation: r'2(t) enter Server 1  q 1 ( t )  0 (see Eq. (2))  q1(t) 
11

12. Overload Controller Design
g
Upstream Server 1 can process all arrival messages without any delay
• before the overload is propagated from its downstream Server 2
()
()
() ()(
p
)
• 2(t)=1(t) and r2(t)=r'2(t) (see previous slide #11)
Queuing dynamics of Server 2
Queuing delay of Server 2
g
y


q2 (t )  1 (t )  r2 (t )   2 (t )   2 (t )
(3)

2 (t )  [r2 (t )  1 (t )  2 (t )  2 (t )] / 2 (t ) (4)
( )
• Each request message corresponds to a response message [SIP RFC]
• Thus request message service rate (i.e., the response message rate
1(t)) can approximate the total service rate 2(t)


Queuing delay of Server 2  2 (t )  [r2 (t )  1 (t )   2 (t )  1 (t )] / 1 (t ) (5)
• Round trip delay of upstream server can approximate queuing delay of
overloaded downstream server 2(t)
 when overload happens and queuing delay is dominant
PI controller regulates retransmission rate r'2(t)
t

r2 (t )  K P e(t )  K I 0 e( )d
t
 K P ( 0   2 (t ))  K I 0 ( 0   2 ( ))d
12

13. Feedback Overload Control System
Figure 4.Block diagram of feedback SIP overload control system
g
g
y
Control plant P(s)=2(s)/r'2(s)= {2(t)}/ {r'2(t)}1/(1s)
PI controller C(s)=KP+KI/s
( )
Open-loop overload control system G(s)=C(s)P(s)=(KP+KI/s)/(1s)
Positive phase margin m of G(s) can guarantee control system stability
PI controller gains can be obtained based on phase margin m
KP 
1 tan( m )
1  tan 2 ( m )
KI 
1
1  tan 2 ( m )
13

14. SIP Overload Control Algorithm (RTDC)
SIP Overload Control Algorithm (RTDC)
14

15. k
l
l
SIP Network Topology For OPNET Simulation
15

16. Scenario to Validate Overload Control Algorithm
• Poisson distributed message generation rate and service rate
• Two typical overload scenarios
• 4 originating servers generated original messages with the same rate
o= (1/4)1; Mean message arrive rate of Server 1 was 1=4o
• Mean service capacity of each originating server was Co=500 messages/sec
Scenario 1
Initial overload at
Server 1 due to
demand burst
• Mean arrival rate 1=800 messages/sec (emulating a
800
short surge of user demands) from time t=0s to t=30s
• Mean arrival rate 1=200 messages/sec (emulating
regular user demands) from time t=30s to t=90s
t 30s t 90s
• Mean service capacities of two proxy servers were
C1=C2=1000 messages/sec
Scenario 2
S
i
Initial overload at
Server 2 due to
server slowdown
• M
Mean arrival rate 1=200 messages/sec
i l t
200
/
• Mean server capacity C1=1000 messages/sec
• Mean server capacity C2=100 messages/sec (emulating
100
server slowdown) from time t=0s to t=30s, and C2=1000
messages/sec from time t=30s to t=90s
16

17. Simulation Results of Scenario 1
Simulation Results of Scenario 1
7
NOLC Queu size q0 (messages)
ue
1
Queue size q (messages
s)
12000
x 10
NOLC q1
OLC q1
6
5
4
3
2
1
0
0
10
20
30
40
50
60
70
80
90
Time (sec)
Queue size q1 (messages) of Server 1
versus time
10
NOLC q0
OLC q0
8
9000
6
6000
4
3000
2
0
0
10
20
30
40
50
60
70
80
OLC Queue size q0 (messa
e
ages)
4
8
0
90
Time (sec)
Queue size qo (messages) of an originating
server versus time
• Without overload control algorithm applied, Server 1 became CPU overloaded
 overload deteriorated as time evolves, leading to eventual crash of Server 1
• Overload control algorithm made queue size of Server 1 increase slowly
 taking 27s to cancel the overload at Server 1 after new user demand rate
reduced at time t=30s
17
 11s faster than RRRC algorithm proposed by IEEE Globecom 2010

18. Simulation Results of Scenario 2
4
x 10
10
6
2
4
1
2
0
0
10
20
30
40
50
60
70
80
0
90
Time (sec)
Queue size q1 (
Q
i
(messages) of Server 1
) fS
versus time
x 10
1.8
2
3
2
Queu size q (messag
ue
ges)
8
OLC q1
1
NOLC q1
4
4
OLC Queue size q (me
Q
essages)
1
NOLC Queue size q (m
Q
messages)
5
NOLC q2
1.6
OLC q2
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
10
20
30
40
50
60
70
80
90
Time (sec)
Queue size q2 (messages) of Server 2
versus time
• Without overload control algorithm applied, overload was propagated from Server
2 to Server 1 when initial overload h
t S
h i iti l
l d happened at S
d t Server 2
• Persisted overload would crash Server 1 after Server 2 resumed its normal
service
O e oad control algorithm p e e t o e oad p opagat o to Se e 1
prevent overload propagation Server
• Overload co t o a go t
 taking only 7s to cancel the overload at Server 2
 2s faster than RRRC algorithm proposed by IEEE Globecom 2010
18

19. Conclusions
 Employing control-theoretic approach to
p
 model SIP overload problem as a feedback control p
problem
 Developing Round Trip Delay Control (RTDC) algorithm
(a PI rate control algorithm) to mitigate the overload by
 controlling retransmission rate
t lli
t
i i
t
 claiming round trip delay below desirable target value
 Simulation results demonstrate that RTDC (implicit SIP
( p
overload control) can
 prevent the overload propagation
 cancel the overload effectively
 Our solution does NOT require modification in the SIP
header and time-consuming standardization process
 can be freely implemented in any SIP servers of different carriers
19

20. Remarks (1)
 Explicit SIP overload control algorithm requires the modification in the
SIP header and the cooperation among different carriers in different
countries
 Implicit SIP overload control algorithm does NOT require the
modification in the SIP header and the cooperation among different
carriers in different countries. Any carrier can freely implement implicit
SIP overload control algorithm in its SIP servers to avoid potential
widespread server crash
 OPNET simulation code f 3 implicit SIP overload control algorithms
i l ti
d for i li it
l d
t l l ith
(RRRC, RTDC, and RTQC) published by IEEE Globecom 2010/ICC
2011 available for non-commercial research use upon request
 RTDC algorithm (proposed by this IEEE ICC 2011 paper) has been
recommended as White paper by TechRepublic (an online trade publication
and social community for IT professionals, part of the CBS Interactive)
http://www.techrepublic.com/whitepapers/design-of-a-pi-rate-controller-formitigating-sip-overload/25142469
20

21. Remarks (2)
 Journal version discusses how to apply RTDC algorithm to mitigate SIP
overload for both SIP over UDP and SIP over TCP (with TLS)
 “Applying control theoretic approach to mitigate SIP overload,”
y
( )
Telecommunication Systems, 54(4), 2013, pp. 387-404. Available at
http://www.researchgate.net/publication/257667871_Applying_control_theoretic
_approach_to_mitigate_SIP_overload
 Survey on SIP overload control algorithms: “A Comparative Study of SIP
y
g
p
y
Overload Control Algorithms,” Network and Traffic Engineering in Emerging
Distributed Computing Applications, IGI Global, 2012, pp. 1-20.
http://www.igi-global.com/chapter/comparative-study-sip-overloadcontrol/67496
t l/67496
http://www.researchgate.net/publication/231609451_A_Comparative_Study_of
_SIP_Overload_Control_Algorithms
 Discussions on control system design can be found in the answers to the
ResearchGate question “What are trends in control theory and its
applications in physical systems (from a research point of view)? ”
https://www.researchgate.net/post/What_are_trends_in_control_theory_and_its
https://www researchgate net/post/What are trends in control theory and its
_applications_in_physical_systems_from_a_research_point_of_view2
21