- Worked under Professor Dragan Hrnjez, on project in Voice Over IP domain. - Analyzing and Optimizing an Enterprise Network to Support Voice traffic using Opnet. - Involves working on SIP signaling protocol and different voice codecs.

1. Enterprise Network - VoIP Service Design
and Analysis
TCOM 631- Voice Over IP
Presented by :
• Chinmay Upasani - G00935325
Presenting to Juniper Networks

2. Project Outline -
• Project Goals
• Constrains and Assumptions
• Codecs Optimization
• Upgrading VoIP considering single points of failure
• Discrete Event Simulation with SIP Proxy Server, Callers and Callees
• Adding PSTN Gateways on the Proxy Server side
• Future Scope

3. Network Topology
Figure 1.1: OPNET network model provided.

4. Project Goals -
• This project analyses an existing enterprise network to make it able to support Voice over IP
(VoIP) requirements.
• Re-engineer the given network to make both real-time voice and data share a common IP
infrastructure instead of a traditional TDM-based voice network.
• PART I – FLOW ANALYSIS -
• Analyzing and compares the following codecs: G.711, G.711 (with silence suppression), G.729 A,
G.729 A (with silence suppression), G.729 and G.728 16K (with silence suppression) to satisfy
given constraints
• PART II - DISCRETE EVENT SIMULATION
• Checking the network performance for simultaneous calls to placed between any two
locations in the network and selecting the appropriate codec for further part.
• PART III – ADDING PSTN GATEWAYS
• Adding PSTN traffic to the existing traffic to check its impact on the network
performance.

5. Constrains and Assumptions -
• PART I - FLOW ANALYSIS
• 4.0 or higher MOS for more than 99% of the total voice traffic
• Link utilization on any of the interconnecting links should not be above 85%
• Grade of Service of 99% should be met
• Delay for VoIP e2e bearer traffic <= 80ms (SLA requirement even in case of single network component
failure)
• PART II - DISCRETE EVENT SIMULATION
• Selecting two non-adjacent locations for placing simultaneous calls.
• There will be 16 number of callers and callees who will place calls for 30 minutes.
• PART III – ADDING PSTN GATEWAYS
• The traffic added by PSTN should be equal to 80 % of the existing voice traffic.
• The Codec selected will have minimal changes to the current enterprise network to
support voice.

6. Part I - Network Model with Data traffic
Figure 1.2: OPNET network model with data traffic.

7. Network Analysis and Design with Voice and
Data traffic -
G.711 Codec
Figure 1.3: OPNET network model with voice and
data traffic using G.711 Codec.
Figure 1.4: Flow analysis with voice and data traffic using G.711
Codec

8. Comparing metrics for Voice and
data traffic -
Codec Max Link
Utilization
(%)
Max
Delay
(ms)
MOS # of over utilized
links
G.711 396 119.166 1.22 18
G.711 (silence ) 241 114.27 1.73 11
G.729 A 216 111.92 1.92 10
G.729 A (silence) 151 106.01 2.47 5
G.728 16k 241 114.27 1.73 11
G.728 16k (silence) 164 108.55 2.33 6
Table 1: Performance metrics after importing voice traffic

9. Optimizing G.729 A (with Silence Suppression)
• 7-changes in total configured on this codec to meet the required criteria.
• The applied changes on the topology:
• Upgrading link between Chicago to Columbus from
DS3 to OC3.
• Modification of OSPF cost metric
DC to New York
• Modification of OSPF cost metric
Columbus to Boston
• Upgrading link between Columbus to
Washington DC from DS3 to OC3
• Addition of Link between Seattle to Chicago
• Addition of link between San Antonio and
Washington DC
• Upgrading link between San Antonio to Washington DC from DS3 to OC3
Figure 1.5: OPNET network model with voice and data traffic
using G.729 A (silence)

10. Final optimized results -
Metric Before Changing After Changing
MOS 1.94 4.37
Max Average Delay 111.926 72.777
Network Utilization 216 74
Over utilized link 10 0
Table 2: Final optimized results for G.729A silence suppression

11. Optimizing G.729 A
• 11-changes in total configured on this codec to meet the required criteria.
• The topology used over (G.729 A with silence suppression)
• The applied changes on the topology :
• Bandwidth change to OC3 on the following links
• San Antonio to Oklahoma
• Washington DC to Philadelphia
• Washington DC to Raleigh
• Seattle to Chicago
Figure 1.6: OPNET network model using G.729 A Codec.

12. Final optimized results -
Metric Before Changing After
MOS 3.89 4.37
Max Avg Delay 73.736 72.813
Network Utilization 107 80.2
Over utilized link 3 0
Table 3: Final optimized results for G.729A

13. Optimizing G.728 A (16k Silence Suppression)
Figure 1.7: OPNET network model after running G.728 (16K
Silence Suppression Codec)
• No changes required to the G.729
A Base model.

14. Final optimized results -
Metric Before Changing After changing
MOS 2.33 4.37
Max Avg Delay 108.551 72.794
Network Utilization 164 80.5
Over utilized link 6 0
Table 4: Final optimized results for G.728A 16k silence suppression

15. Performance metrics of optimized networks -
Codec Max Link
Utilization
(%)
Max
Delay
(ms)
MOS # of over
utilized links
G.729 A (silence) 74 72.77 4.37 0
G.729 A 80.2 72.81 4.37 0
G.728 16k (silence) 80.5 72.79 4.37 0
Table 5: Final optimized results for all codecs

16. Link Failures Analysis -
• Three similar failures applied on all codecs :
• Nashville node-turned down
• Link between Columbus to DC-turned down
• Philadelphia node-turned down

17. G.729.A (with Silence Suppression)
failures analysis -
Table 6: Failure analysis for G.729A silence suppression

18. G.729.A failures analysis -
Table 7: Failure analysis for G.729A

19. G.729.A ( 16K Silence Suppression)
failures analysis -
Table 8: Failure analysis for G.728A 16K silence suppression

20. PART II - DISCRETE EVENT SIMULATION
• Creating SIP Proxy Server and selecting endpoints
• Creating Profile Configuration
• Creating Application Configuration
• Creating Subnets for Callers and Callees
• Running DES

21. Selecting SIP endpoints -
Erlang B calculation

22. SIP Server Configuration -
• We selected Chicago as a location for SIP proxy Server.
Figure 2.1: SIP Proxy Server Attributes

23. Profile Configuration -
• Configuring 16 caller and callee profiles.
Figure 2.2: SIP caller Attributes Figure 2.3: SIP callee Attributes

24. Application Configuration -
• Call duration - exponentially distributed with the 4 minute
mean
• Inter-repetition call time – exponentially distributed with the 2
minute mean
• Simultaneous Operation Mode – Start Time uniformly
distributed with min=60, max=70
Figure 2.4: Attribute configuration

25. Creating Subnets -
Figure 2.5: 16 Callees connected to New
York
Figure 2.6: 16 Callers connected to
Atlanta

26. Graph Showing average MOS Value, Number of Calls Setup,
Calls Connected and Calls Rejected
Figure 2.7: Graph Showing average MOS Value, Active calls, Call Duration,
Number of Calls Setup, Calls Connected and Calls Rejected

27. SIP Call statistics and MOS value
for different codecs -
Codec Calls Setup Calls Connected Calls Rejected MOS Value
G.729 A (silence)
128 Feed
16 14 2 3.67
G.729 A (silence)
256 Feed
16 13 3 4.02
G.728 16k
(silence) 128
Feed
16 13 3 4.35
G.728 16k
(silence) 256
Feed
17 14 3 4.36
Table 9: Call statistics for different codecs

28. PART III - ADDING PSTN GATEWAYS ON THE
PROXY SERVER SIDE -
• IN-OUT traffic at the SIP Proxy Server where we are attaching our PSTN gateway
Locations Capacity used
(Mbps)
% Contribution to
total traffic
Capacity used
(Mbps)
% Contribution to
total traffic
Oklahoma 11.06 11.06/138.08 =8 12.88 12.88/141.7=9.08
Minneapolis 36.25 36.25/138.08
=26.25
34.74 34.74/141.7=24.51
Seattle 27.65 27.65/138.08 =
20.02
29.36 29.36/141.7=20.71
Columbus 37.57 37.57/138.08 =
27.21
37.07 37.07/141.7=26.16
Nashville 25.55 25.55/138.08 =
18.50
27.65 27.65/141.7=19.51
IN Traffic OUT Traffic
Table 10: Existing voice traffic analysis (IN and OUT)

29. Adding PSTN Traffic -
• Adding 80% of current voice traffic -
• Total current voice traffic in network = 4581.794 Erlangs
• Calculating 80% of current voice traffic = 3665.4352 Erlangs
• Total 3685 trunks channels for PSTN Traffic

30. PSTN Traffic bandwidth calculation -
Now, for G.728 A (16 k Silence Suppression), the Sample size = 60 bytes
 IP header = 20 bytes
 UDP header= 8 bytes
 RTP header = 12 bytes
So total packet size = 60+40 = 100 bytes.
100 bytes/ packet x 33.33 pps x 8 = 26.664 kbps
Total Voice Bandwidth due to PSTN gateway: 3685 x 26.66
kbps = 95.93 Mbps

31. Link loading due to PSTN traffic
Links PSTN traffic (Mbps) Total traffic (Mbps) Utilization (%)
Chicago to Oklahoma 0.0908 * 95.93 =8.71 12.88+8.71=21.59 21.59/44.74=48.25
Chicago to Minneapolis 0.2451* 95.93=23.51 34.74+23.51=58.25 58.25/148.61=39.19
Chicago to Seattle 0.2071*95.93=19.8 29.36+19.86=49.22 49.22/148.61=33.12
Chicago to Columbus 0.2616*95.93=25.09 37.07+25.09=62.16 62.16/148.61=41.82
Chicago to Nashville 0.1951*95.93=18.7 27.65+18.70=46.35 46.35/148.61=31.19
Links PSTN traffic (Mbps) Total traffic (Mbps) Utilization (%)
Okhlahoma to Chicago 0.08 * 95.93 =7.67 11.06+7.67=18.73 18.73/44.74=41.86
Minneapolis to Chicago 0.2625 * 95.93=25.18 36.25+25.18=61.43 61.43/148.61=41.34
Seattle to Chicago 0.2002*95.93=19.21 27.65+19.21=46.86 46.86/148.65=31.52
Columbus to Chicago 0.2721*95.93=25.29 37.57+25.29=62.86 62.86/148.61=42.29
Nashville to Chicago 0.1850*95.93=17.75 25.55+17.75=43.3 43.3/148.65=29.13
Table 11: Link loading statistics due to PSTN traffic (IN and OUT)

32. Link changes after adding PSTN traffic -
• Nashville – Chicago link upgrade from DS3 TO OC3
• Seattle – Chicago link upgrade from DS3 TO OC3
Figure 3.1: OPNET network model optimized

33. Flow analysis and MOS of final
optimized network -
Figure 3.2: Flow Analysis of Final optimized network
Figure 3.3: MOS of Final Optimized network

34. Enterprise network after inducing voice traffic
(G.728-16k-Silence)
Figure 4.1: Enterprise network after inducing voice traffic

35. Setting up a tunnels between Denver
and Minneapolis -
Figure 4.2: Setting up Tunnel 1 and Tunnel 2 between Denver and Minneapolis

36. Load Balancing the traffic from Denver to
Minneapolis between tunnel 1 and tunnel 2 -
Figure 4.3: Load balancing between Tunnel 1 and Tunnel 2

37. Incoming and outgoing traffic in tunnel 1
and tunnel 2
Figure 4.4: Setting up incoming and outgoing tunnels

38. Thank You