Transcript WPB1: Pa2PaGW Kick

Broadband Europe 2007 We3A.4
Document: Emulation and Simulation Tool for
Design and Optimization of IMS based FMC Networks
Date:
2007-12-05
Originator: Stefanie Braun, Stefan Wahl
Muse confidential
Agenda
>
Motivation of the Tool
>
Emulation and Simulation Tool
>
Usage and Output Results
>
Tool Appliances
>
Conclusions
Broadband Europe 2007 We3A.4 — 2
Motivation
>
Questions on Optimization and Deployment Alternatives for
an IMS System
•
What is the most efficient architecture of an IMS system?
–
•
central  any flavor between central/distributed  distributed
How does the efficiency vary with the deployment and service
scenarios?
size of the network
– type of the services
– the service usage and use case scenarios
–
•
Which function grouping improves the efficiency?
–
•
grouping of functions may simplify the interface  less processing
effort
What is the best deployment strategy of (new) services on the
architecture?
 Development of Simulation and Emulation Tool
Broadband Europe 2007 We3A.4 — 3
Emulation and Simulation Tool
>
Emulation part used to
•
emulate functions and interfaces at
application, call control and media transport layer
– with different (selectable) abstraction levels whereby
– each invocation of functions/interfaces adds a computational effort share
–
•
>
validate the interworking of functions and interfaces
Simulator part
•
•
applies the emulated functions and interfaces
adds user agents and a traffic generator which
–
•
•
•
produce a mixtures of determinable services at offered loads
accumulates the computational efforts
provides statistics on average/peak/hot spot computational
efforts
effort analysis on function, node and architecture level
Objective: Performance comparison, validation and verification
of IMS
like— 4architectures and service scenarios.
Broadband Europe
2007 We3A.4
Emulation and Simulation Tool
Architecture composition and output
>
Graphical or XML based
network architecture
composition:
• Provides means to
configure easily the
network size and the
amount of network
domains
…
• Network element (NE)
are composed by
…
application, call control
and media layer
functions
NE1
• Assignment of a number
of User Agents (UA) to
UA 1
the NE for generation
and termination of
service calls
Broadband Europe 2007 We3A.4 — 5
NEm
NE5
NE4
AS
mobility logic
S-CSCF
…
NE2
I-CSCF
P-CSCF
UA 2
NE3
MRFP
…
UA
UA
Traffic generator
UA
UA n
Emulation and Simulation Tool
Architecture composition and output
Graphical output of
the simulator part
NE2
peak or average
computational effort
S-CSCF
…
NE3
e.g.
UA 2
setup video
call to UAx
NE1
NE2
NE3
NE4
…
NE5
NEm
per function or
component
Processing Effort per
Component
MRFP
…
UA
release audio
call
UA
UAx
Traffic generator requests and releases independently
network services and end2end calls
Broadband Europe 2007 We3A.4 — 6
UA n
setup teleVoting
call
Traffic Generator
90
80
70
60
50
40
30
20
10
0
Effort
..
P-CSCF
NEm
Effort
AS
…
per network element
800
700
600
500
400
300
200
100
0
I-CSCF
UA 1
Processing Effort of Network
Elements (NE)
mobility logic
Effort
NE1
AS
PCS
CF
I-C
SC
F
SCS
CF
M
RF
M
ob
P
ilit
yL
og
ic
…
Components
per function
or component
over time
Time Dependent Effort for
S-CSCF
40
30
20
Effort
10
0
0.00 4.00
8.00 12.00
12.00 16.00
Time
16.00 20.00
20.00 24.00
Research of Processing Effort Values
Input parameters for the emulation/simulation tool
Extraction of SIP processing effort values
from performed tests and literature*
•
•
Parsing

Consumes about 25% of total message processing time

Parsing time grows linearly with message size
String and Memory Manipulation


Proprietary string operations and less processing stages reduce the impact on the total
processing message time from 45% to 7%
Proprietary memory operations reduce processing message time from 18
to 3 %
Parsing + string handling + memory allocation:
take 33 - 83 % of total message processing time in SIP proxies
Inter NE function (e.g. CSCF) interfaces could replace SIP protocol
Replacing SIP protocol with more efficient inter NE protocols (e.g. binary, ordered)
improves the message processing time
* Articles :
„On
SIP Performance“,
Cortes,
Ensor,
Broadband
Europe 2007
We3A.4
— 7Esteban, Bell Labs Technical Journal 2004
„Measurement of the SIP Parsing Performance in the SIP Express Router“, S. Wanke et al., Proc. EUNICE 2007
Emulation and Simulation Tool
Examples for architecture optimization
>
Performance gains due to grouping of IMS functions
>
Architecture and protocol investigations based on
computational processing effort
•
•
•
>
Scenario A: Call sessions with media processing
Scenario B: TeleVoting with high signaling, AS and media load
Scenario C: Trend analysis – increasing amount of AS services
Comparison between centralized and distributed AS server
concept
Broadband Europe 2007 We3A.4 — 8
Emulation and Simulation Tool
Current IMS
Performance gains due to grouping of IMS functions
Standard IMS case
visited
for multi-domain with roaming A
home A
S
I
I-CSCF
P
Average
= 6.2 CSCF hops
P-CSCF
UE
UE
P
home B
I
I
S
visited
B
I
S-CSCF
P
P
UE
Home-Home: 6 CSCF Instances
UE
Foreign-Foreign: 8 CSCF Instances
MS-ER Solution
Terminal / Peering
Grouping IMS CSC-Functions in MS-ER
PSI
UE
I
Typical Example
3.5
43 %
SP
UE
Broadband Europe 2007 We3A.4 — 9
Probability Histogram
of combined CSC-Functions
- 10 IMS domains
- Calls equally distributed
across all domains
- 20% users in roaming
Most advantageous CSCF combinations:
SP-CSCF, PSI-CSCF, SI-CSCF, …
Average CSCF hops:
Reduction:
Simulation Result:
I-CSCF
(S-P)-CSCF
(P-S-I)-CSCF
P-CSCF
(S-I)-CSCF
(P-S)-CSCF
About 57 % of the
simulated cases
require 3 CSCF hops
Emulation and Simulation Tool
Examples for architecture optimization
>
Performance gains due to grouping of IMS functions
>
Architecture and protocol investigations based on
computational processing effort
•
•
•
>
Scenario A: Call sessions with media processing
Scenario B: TeleVoting with high signaling, AS and media load
Scenario C: Trend analysis – increasing amount of AS services
Comparison between centralized and distributed AS server
concept
Broadband Europe 2007 We3A.4 — 10
Emulation and Simulation Tool
Architecture investigations: Call sessions with media processing
Architecture Comparison
...
S
...
S
P
...
MRF
UE
UE
Test case: Typical IMS
MS-ER
intra
MS-ER
protocol
P
intra ...
S
AS
MRF
3GPP SIP
MS-ER
AS
S
inter
intra ... MS-ER
protocol
>
2 independent administrative domains
>
Clients are equally distributed across
domains
>
No roaming clients
>
Clients establishes calls randomly
>
80% standard calls
>
20% calls use transcoding service
>
IMS SIP stack processing effort: 100%
>
Processing effort values for the inter and
intra MS-ER protocol are varied
3GPP SIP
...
P
3GPP SIP
Home network
AS
...
IMS
Scenario Description
MRF
UE
Test case: MS-ER
Broadband Europe 2007 We3A.4 — 11
intra ...
P
3GPP SIP
UE
•
Intra/Inter: 20% / 40%
•
Intra/Inter: 50% / 70%
Emulation and Simulation Tool
Architecture investigations: Call sessions with media processing
Test Results
Total Average Effort
Effort Estimation
120
Effort [%] _
100
100%
-28%
80
Total Peak Effort
-17%
-17%
-4%
60
40
20
0
IMS
3GPP SIP
Typical IMS
implementation with SIP
based interfaces
Broadband Europe 2007 We3A.4 — 12
MS-ER
intra= 20%
inter= 40%
of 3GPP SIP
MS-ER
intra= 50%
inter= 70%
of 3GPP SIP
MS-ER implementation
with optimized “inter”
and “intra” interfaces
Emulation and Simulation Tool
Examples for architecture optimization
>
Performance gains due to grouping of IMS functions
>
Architecture and protocol investigations based on
computational processing effort
•
•
•
>
Scenario A: Call sessions with media processing
Scenario B: TeleVoting with high signaling, AS and media load
Scenario C: Trend analysis – increasing amount of AS services
Comparison between centralized and distributed AS server
concept
Broadband Europe 2007 We3A.4 — 13
Emulation and Simulation Tool
Architecture investigations: TeleVoting scenario
Test Results
Televoting
Basic Calls
Effort Estimation
120
...
3GPP SIP
AS
...
S
CT Vote App.
...
P
100
100%
-28% -28%
80
Total Peak Effort
-17% -16%
60
40
20
S
0
IMS
3GPP SIP
...
Home network
Central App. Server
UE
Effort [%] _
Architecture Comparison
Total Average Effort
MRF
MS-ER
intra= 20%
inter= 40%
of 3GPP SIP
MS-ER
intra= 50%
inter= 70%
of 3GPP SIP
Home network
MS-ER 1
intra MS-ER
protocol
P
intra ..
S
AS
Vote result
CT
Vote
App.
intra ..
MRF
3GPP SIP
UE
Co-located App. Server
Broadband Europe 2007 We3A.4 — 14
Higher efficiency for co-locating
call control, application layer and media
transport functions to one MS-ER node
• Improved efficiency also for heavy load
generating TeleVoting services
•
Emulation and Simulation Tool
Examples for architecture optimization
>
Performance gains due to grouping of IMS functions
>
Architecture and protocol investigations based on
computational processing effort
•
•
•
>
Scenario A: Call sessions with media processing
Scenario B: TeleVoting with high signaling, AS and media load
Scenario C: Trend analysis – increasing amount of AS services
Comparison between centralized and distributed AS server
concept
Broadband Europe 2007 We3A.4 — 15
Emulation and Simulation Tool
Architecture investigations: Increasing amount of AS services
Test Cases
MS-ER 1
intra MS-ER
protocol
S
AS
3GPP SIP
S
inter
MS-ER
protocol
MS-ER + central AS
Peak Effort Estimation
MS-ER 2
120
100
intra ..
P
Effort [%] _
central AS
Test Results
P
3GPP SIP
3GPP SIP
UE
intra ..
P
S
inter
MS-ER
protocol
AS
-17%
80
60
40
intra ..
callFw=
100%
the MS-ER + AS
histogram bars are
used as effort
references
MS-ER 2
S
•
P
3GPP SIP
3GPP SIP
-11% -14%
callFw= 0% callFw= 20% callFw= 40% callFw= 60% callFw= 80%
co-located AS
intra ..
-8%
0
UE
AS
-5%
20
Home network
MS-ER 1
0%
MS-ER
Home network
UE
UE
Broadband Europe 2007 We3A.4 — 16
The efficiency of the MS-ER concept increases
with the increasing number of services
applying AS functions
Emulation and Simulation Tool
Examples for architecture optimization
>
Performance gains due to grouping of IMS functions
>
Architecture and protocol investigations based on
computational processing effort
•
•
•
>
Scenario A: Call sessions with media processing
Scenario B: TeleVoting with high signaling, AS and media load
Scenario C: Trend analysis – increasing amount of AS services
Comparison between centralized and distributed AS server
concept
Broadband Europe 2007 We3A.4 — 17
Emulation and Simulation Tool
Comparison co-located  central AS
MS-ER architecture with co-located AS
MS-ER 1
intra ..
S
P
intra ..
AS
AS
inter
MS-ER
protocol
MS-ER 2
intra ..
S
intra ..
P
3GPP SIP
Break Event Points for co-located AS
MS-ER + central AS
MS-ER + co-located AS
Peak Computational Effort
Estimation
3GPP SIP
Home network
MS-ER central AS
is more efficient
Effort [%]
MS-ER architecture with central AS
MS-ER 3
AS
MS-ER co-located AS
is more efficient
20 MS-ER
inter MS-ER protocol
MS-ER 1
S
P
intra ..
AS
AS
inter
MS-ER
protocol
MS-ER 2
S
3GPP SIP
≈
16
P
3GPP SIP
≈
4 MS-ER
20
24
28
32
36
40
Intra MS-ER protocol effort [%]
Parameter: Inter MS-ER protocol effort = 40 %
Home network
To maintain the peak effort gain for an increasing number of MS-ER nodes, the co-located AS
MS-ER architecture requires a more performing intra MS-ER communication.
Broadband Europe 2007 We3A.4 — 18
Emulation and Simulation Tool
Further use cases
>
Resource Manager
•
•
>
Network planning
•
•
>
hardware, service-, signaling-, and media software components
algorithm development
optimization of existing network (identification of hot spots)
extension of the network
Distributed service execution
•
evaluation of more advanced strategies
Cross
domain
session
mobility
L3 Mobility
Domain 1
IMS Core
Domain
2
AS
AS
Session Signaling
Media Processing
Stream Switching
IP Transport
Service logic
for sea mless
session
mobility
control
Service logic
for joint
mobility
detection
Distributed
Distributed AS
AS and
and MRF
MRF
enable
enable cross
cross domain
domain
seamless
seamless session
session mobility
mobility
Broadband Europe 2007 We3A.4 — 19
Foreign
N etwork
AS
AS
Emulation and Simulation Tool
Conclusion
>
Emulation features allow to
•
•
•
>
integrate functions with required abstraction levels
reduce the communication protocols to the relevant functions
compare service introduction strategies into an IMS architecture
Computational effort evaluation to
•
•
•
•
•
compare various variants of IMS network architectures
identify processing effort values and hot spots at different levels:
domains, network node and functions
estimate performance dependencies and critical optimization
parameters of a network architecture
derive the influence of protocol and functional improvements
check the influence of a resource management algorithm
Broadband Europe 2007 We3A.4 — 20
Questions ?
Muse confidential