Performance Analysis of VOIP over GRE Tunnel

Published Online November 2015 in MECS DOI: 10.5815/ijcnis.2015.12.01

During recent years we experienced explosive growth of Voice over IP (VOIP) technology. The driving motivation for using VOIP is cost saving, mostly for large companies with pre-existing network infrastructure. Another advantage of carrying voice traffic over IP networks is ability to integrate voice and data in one application [3]. However, transmitting voice over a public IP-based network such as internet has raised many security issues. In the case of VOIP, security concerns can be mutual authentication of called and caller for voice protection, session privacy and encryption (to avoid eavesdropping, sniffing, and man in middle attack), session integrity (to avoid altering VOIP packets), and protection of related data to a VOIP connection such as packets pertaining to billing system. Also, Denials of Service (DoS) attacks in VOIP applications try to bring down the system by over flooding or making it busy by sending large number of call requests [4]. Another security issue is VOIP spam which is similar to Email spam. In this case spammers will overflow voice mail inboxes with unwanted message.

Some techniques have been proposed to handle security issues in IP based communication, but among them employing Virtual Private Network (VPN) is more widely used [7]. However, VPN is expected to have an impact (a negative or positive) on VOIP performance. Other important factors that can affect VOIP service quality are various signaling protocols such as H323, and SIP; audio codecs used such as G.711, G.729, G.728, G.726, etc., and related encoding algorithms such as PCM (Pulse Code Modulation), Adaptive Differential PCM, etc. VPN protocols that can be employed are PPTP, L2TP, IPsec and GRE. Clearly, depending on the choice of VPN tunneling protocols, type of services provided to VOIP packets- by VPN- will differ and therefore each VPN protocol is supposed to have a different impact on VOIP performance. Some prior research papers have investigated impact of signaling protocol and/or choice of audio CODECs on quality of VOIP service. In this paper we examined role of VPN protocols as one of QoS parameters in VOIP a communication. We used OPNET to simulate the behavior of a VOIP system while running over IP VPN tunnel.

• II.    Literature Review

• A.    Related Works

  Telecommunication standardization bodies such as International Telecommunications Union (ITU-T) and several researchers have outlined a number of contributing factors in Quality of Service (QoS) for a voice connection. These factors include ITU-U codecs and algorithms, end-to-end delay, jitter (also known as delay variation), packet loss, and network design [2, 1112]. According to ITU-U guidelines, a voice call facing a delay greater than 150 ms (note: some authors refer to 200 ms) and/or a jitter of greater than 20 ms is not considered to be of a good quality, and accordingly any voice call facing delay of greater than 300 ms and/or a jitter of greater than 50 ms is considered to be of a poor quality [13].

Table below outlines the accepted voice quality measures.

Table 1. ITU-T VoIP Quality Measurement [13]

Network parameter	Good	Acceptable	Poor
Delay (ms)	0-150	150-300	> 300
Jitter (ms)	0-20	20-50	> 50

There are a numbers of related studies in the field which investigated performance of VOIP calls in different network environments along with different signaling protocols, voice codecs, QoS parameters and security protocols. Muhamad Amin [11] conducted a study with regards to three aspects of VOIP communications, namely call signaling protocols (H323 and SIP), VPN protocols, and network environment (Ethernet and WLAN). From the security prospective authors used VPN with two distinct tunneling protocols namely IPsec and PPTP. According to the results congestion was found to negatively impact voice quality parameters such as delay and jitter. VPN reported to have a similar effect on voice traffic. In a non-ideal network environment, the voice quality parameters even showed worse results compared with an ideal network environment.

Another interesting study in the field is presented by Ibrahim S. I. Alsukayti and Timothy J. Dennis [7]. They compared performance of VOIP while running over BGP/MPLS VPN network with that of VOIP while running MPLS network. BGP/MPLS VPN is a VPN technology which integrates Multiprotocol Label Switching (MPLS) features with security aspects of VPN. Results of the paper suggested that not only a VPN over BGP/MPLS has a positive impact on VOIP quality but also positively improves performance of VOIP as compared with its performance over an MPLS network. Regarding different VOIP codecs, a comparison has was done between G.711, G.723.1 and G.729A, over the BGP/MPLS VPN network model and the comparison result showed that G.729A (bit rate = 8 kb/s) is the best choice of voice codec for such a scenario (i.e. BGP/MPLS VPN) due to bringing a balance between end-to end delay and bandwidth efficiency.

Gouda I. Salama et al., [2] examined “impact of IPsec on the quality of transmitting voice over communication links using OPNET simulator”. Result of their research showed the IPsec results in an increase in packet loss, end to end delay, call setup time, and jitter.

Barbieri et al., [10] proposed to reduce size of IPsec encapsulated packets (actual IP packet inside IPsec header) by almost 4 bytes using compression. This will address Quality of Service for IPsec transmitted packets over IP network. This approach, whoever, is criticized by Gouda I.Salama et al., [10] for neglecting actual compression time, which in turn will lead to a processing delay that might be even larger than encryption delay.

Shankar R. R [1] performed a comparison among three commonly used VOIP codecs used in peer to peer VOIP networks and found that G.729 is a better choice for VOIP applications, because it requires lower bandwidth as compared to G.711 and G.723. In another study by Henning and Jonathan Rosenberg [14], a comparative examination of the services, complexity, extensibility and scalability of the two protocols: SIP and H.323 was performed. Findings of the study suggested that SIP and H.323 provides similar services, but SIP showed less complexity, and better extensibility and scalability [14].

Last but not least, in a study performed by A. Asadi Eskandar, Mahbubur. R. Syed , and M.B. Zarei [15], impact of parameters arising from network in performance of SIP over IP VPN was examined. According to the study, VPN doesn’t necessary brings the VOIP performance. For example VPN showed no negative impact in call setup time if SIP proxy servers are located in the same network segment as end-point (caller and callee) SIP phones are located. Also, findings of the study suggested that use of VPN in combination with a proper design (in the case of the above mentioned experiment, proper placement of SIP Proxy Servers) can actually improve the VOIP performance [15].

• B.    Voice Codec

CODEC (Coder/Decoder) is one of the essential components of VOIP. At the sender side, CODECs converts analogue voice signals to digital signals, compresses and encodes to predetermined format. ITU-T introduced and standardized various CODECs. Most commonly used ones are G.711, G.722, G.723.1 and G.729A each working with different bit rate and vary in performance [4], as detailed below:

• •    G.711: Minimum bandwidth needed is 128 kbps

and its speech transmission is precise.

• •    G.722: Different compression is possible

• •    G.723.1: Voice quality is high but consumes high

processor power

• •    G.726: Version of G.723 and G.721

• •    G729: Has efficient utilization of bandwidth.

License required.

Table 2, contains a list of some VOIP’s CODESs and their related bandwidth [1]:

Table 2. VOIP CODECs and Relevant Bandwidth

CODEC	Bandwidth (kbps)
G.711	64
G.722	64
G.723.1	6.3
G.726	32
G.728	16
G.729	8

Mean Opinion Score (MOS) was introduced by ITU-T and represents multimedia quality form user’s prospective, ranged 1 (poor) to 5 (excellent) [4].

Each Codec and its relevent MOS is listed in table 3.

Table 3. MOS Related to each Codec

CODEC	Bit Rate (kbps)	MOS
G.711	64	4.4
G.723.1	6.3	3.9
G.726	32	3.85
G.728	16	3.61
G.729	8	3.92

As is shown in table, speech quality degrades in a nonliner manner with the decrease of data rate [1].

Another component of a VOIP connection is packetization in which encoded voice is encapsulated in packets. Each packet contains different headers at different layers, such as real-time Transport Protocol (RTP), User Datagram Protocol (UDP), and Internet Protocol (IP). Fig. 1 illustrates end to end voice transmission in a VOIP system. After encoding and packetization the packets are send out over IP based network and the process repeats reversely in the destination:

Fig. 1. End-to-End Voice Transmission 1

Playout buffer at the receiver end is used to smooth playout by mitigating the incurred jitter during transmission.

• C.    H.323 Signaling Protocol

One of the main key areas of a VOIP system is signaling protocol which makes different network elements to communicate with each other, establish and terminate calls [5]. Two commonly used signaling protocols are H.323 protocol suite introduced by ITU-T and SIP by ITEF. H.323 is more common standard and operates over packet switched networks such as IP network [3].

The H.323 protocol suit consists of three main control areas:

• •    Registration, Admission, and Status (RAS) signaling protocol, also called H.225 Signaling, is a transaction oriented protocol which operates between a H323 terminal or endpoint and a gateway. An endpoint, with the help of RAS protocol can access to a gatekeeper which has address translations. Endpoint can register or unregister with a gatekeeper [5].

• •    Call control/Call setup: H.225 signaling protocol for call control is used to establish connection between H.323 endpoints.

• •    H.245 media control and transport: H.245

protocol provides logical channel audio, data and video transmission as well as control channel information [6].

• D.    SIP Signaling Protocol

  SIP (Session Initiation Protocol) is one of the VOIP standard peer to peer protocols which is defined in RFC 2543 and standardized by the IETF MMUSIC Working Group. This protocol contains initiation, termination and also modification standards for user sessions which consist of video or audio elements, online games, instant messaging, virtual reality or generally multimedia elements [1].

SIP configuration includes user agent and proxy server. User agent client creates and terminates requests while user agent server, generates responses after receiving SIP requests. The response can be accept, reject or redirect. Proxy servers can also act as client and generate requests on behalf of other clients. Mainly proxy servers act as routers. Fig. 2, shows SIP components and protocols.

Fig. 2. SIP components

• E.    Virtual Private Network (VPN)

The VPNs are responsible for providing a secure connection through a public network like the internet. VPN employs a functionality known as IP tunnel that is a virtual point to point link between two end nodes, which may be located in different networks with number of intermediate networks in between [9].

There are a number of different tunneling protocols such as GRE (Generic Routing Encapsulation), PPTP (Point-to-pint Tunneling Protocol), L2TP (layer 2 Tunneling Protocol), IPsec (IP security) and SSTP (Secure Socket Tunneling Protocol).

GRE is a tunneling protocol developed by Cisco and provides encapsulation of a wide range of network layer protocols inside point to point links. GRE tunnels are normally established between a source and destination router with packets encapsulated with GRE header [8]. Fig. 3 illustrates packet encapsulation in GRE.

Fig. 3. Packet Encapsulation in GRE

• III.    Methodology

For the purpose of this research we employed simulation to conduct the research. Simulation is valid approach to model a real scenario or a system. By simulating a system expected behavior of the system under different conditions can be studied. Simulation is valid approach to model a real scenario or a system. By simulating a system expected behavior of the system under different conditions can be studied. In this study we utilized OPNET Modeler as the simulation tool. OPNET Modeler is powerful simulation tool which has been adopted by researchers (and also in the industry), offering a complete simulation and development environment for simulation and modeling the communication networks. OPNET Modeler provides discrete event simulation feature, which can be used to study the performance and behavior of a model. By utilizing OPNET Modeler’s GUI, different hypothetical scenarios can be configured and studied. For modeling and simulation of a network topology, OPNET Modeler offers three main editors at network level, Node level and Process level. Network-level editor provides high level modeling for deisgn and creation of the network(s) to be studied. Node-level editor can be used to model and define the behvaoir and flow structure of ineteranl modules within a network level componet. Procees-level editors repersents Finite State Machine decsitiptions and C/C++ source codes realted to each state of the process model [16].

In this section we describe the simulation process and the network topology. OPNET Modeler version 17.5 was used as our simulator. Created and configurated Network topology in this work is shown in fig. 4.

The network topology, shown in fig.4 represents three network segments or sites, namely A, B, and C. Each network segment has a router, an access layer (Layer 2) switch. Site-A (bottom-right) and Site-B (bottom-left), each, has one IP phone device (generating VOIP traffic) and one PC (generating data traffic. Site-C (top-center) has either a H.323 Gatekeeper or a SIP-Proxy Server (based on the running signaling protocol in different scenarios) and one server, which will receive and send packets to and from PCs located in site A and B. Each site (network segment) is connected via IP network cloud (resembling Internet). In this work we studied and evaluated impact of different factors such as signaling protocols (i.e. H.323 and SIP), VPN Tunneling protocols (i.e. GRE), and voice codecs (i.e. G.711 and G.723) –as identified by prior research- that play a role in quality of a VOIP communication. There are two general network scenarios: 1) with VPN tunneling and 2) without VPN tunneling. VPN tunneling protocol used in this experiment is GRE. Each of the scenarios is simulated with two different profiles, namely, data and voice. As the names suggest a voice profile was assigned to each IP phone and the data profile was assigned to each PC. Data profile included mix traffic of such applications as Web, Email, Database and File Transfer. IP phones were configured to strictly communicate to each other and PCs were configured to send/receive traffic to/from the Server located in Site-C (top-center).

Fig. 4. VOIP simulated Network Configuration

To exlude impact of scuch factors as network congestion of we chose to configure and generate light traffic pattern during simulation in all scenarios (i.e. with VPN and without VPN). Data traffic was generated and sent in serial order throughout the simulation with unlimited repetitions and applications running simultaneously. In the case of voice traffic, 15 phone calls were generated in serial order; each call had a duation of 180 seconds. Configured profiles for Data traffic are shown in fig. 5 and fig. 6, respectively. Fig. 7 and 8 repserenst voice Codec atributes for G.711 and G.723 Codesc, respectovely. Lastly, fig. 9, represents global configuration and IP addressing scheme where VPN tunnel is established.

= DATA

;-• Profile Name В Applications г Number of Rows № Database Access (Light)	DATA (...) 5
№ Email (Light)
№ Video Conferencing (Light)
R Web Browsing (Light HTTP1.1)
1 Name	Web Browsing (Light HTTP1 1)
| Start Time Offset (seconds)	uniform (5.10)
; Duration (seconds)	End of Last Task
Is Repeatability                                             (...)
I- Inter repetition Time (seconds)	exponential (300)
!- Number of Repetitions	Unlimited
Repetition Pattern	Serial
® Ale Transfer (Light)                                     ...
[■•Operation Mode	Simultaneous
[•• Start Time (seconds)	uniform (100. 110)
;■• Duration (seconds)	End of Last Application
$ Repeatability	Once at Start Time

Fig. 5. Configured Data Profile

Preview

Subnet ID Member Address Member Node Interface Name				Is Router?
1 10.1.2.0/24    10.1.2.1	RouterB	TunnelO	Yes
2 10.1.1.0/24    10.1.1.1	RouterA	TunnelO	Yes
3 192.168.3.0/24 192.168.3.1/24	RouterC	IFO	Yes
4 192.168.1.0/24 192.168.1.10/24	Phonel	IFO	No
5                 192.168.1.11/24	PC1	IFO	No
6 192.168.2.0/24 192.168.2.10/24	Phone2	IFO	No
7                192.168.2.11/24	PC2	IFO	No
8                192.168.2.1/24	RouterB	IFO	Yes
9 192.1.3.0/24 192.1.3.1/24	RouteiC	IF10	Yes
10                192.1.3.2/24	IP CLOUD	IF1	Yes
11 192.168.3.0/24 192.168.3.10/24	GK	IFO	Yes
12                192.168.3.11/24	Server	IFO	No
13 192.168.1.0/24 192.168.1.1/24	RouterA	IFO	Yes
14 192.1.2.0/24 192.1.2.1/24	RouterB	IF10	Yes
15                 192.1.2.2/24	IP.CLOUD	IFO	Yes
16 192.1.1.0/24 192.1.1.1/24	RouterA	IF10	Yes
17                 192.1.1.2/24	IP_CLOUD	IF2	Yes

Fig. 9. Global Configuration – VPN Established

9 Profile Configuration b Number of Rows

19 VOIP

	•• Profile Name	VOIP
1 =	a Applications
b Number of Rows                               1
® Voice over IP Call (PCM Quality)                      ...
	•• Operation Mode	Serial (Ordered)
	•• Start Time (seconds)	constant (60)
	•• Duration (seconds)	End of Last Application

S Repeatability

b Inter-repetition Time (seconds)	constant (60)
b Number of Repetitions	constant (15)
i- Repetition Pattern	Serial

Fig. 6. Configured VOIP Profile

Fig. 7 and Fig. 8 show voice Codec attributes running during simulation.

Attribute	Value
@ Voice Encoder Schemes	(...)
i- Number of Rows	4€
+ PCM                                               ...
I- Codec Type	PCM
Ь Name	G.711 (silence)
Ь Frame Size (secs)	10 msec
Ь Lookahead Size (secs)	0 msec
'■■ DSP Processing Ratio	1.0
i- Coding Rate (bits/sec)	64 Kbps
i- Speech Activity Detection	Enabled
L Equipment Impairment Factor (Ie)	unknown
■■■ Packet Loss Robustness Factor (BpI)	default

Fig. 7. G.711 voice Codec attribute

	■ Codec Type	MP-MLQ
	■Name	G.723.1 6.3K (silence)
	■ Frame Size (secs)	30 msec
	■ Lookahead Size (secs)	7.5 msec
	■ DSP Processing Ratio	ID
	■ Coding Rate (bits/sec)	6.3 Kbps
	■ Speech Activity Detection	Enabled
	■ Equipment Impairment Factor (Ie)	15
	■ Packet Loss Robustness Factor (BpI)	16.1

Fig. 8. G.723 Voice Codec attributes

Fig. 9, represents global configuration and IP addressing scheme where VPN tunnel is established.

• IV.    Results

In the first set of simulation experiment (i.e. Experiment.1HHO: scenario 1.1 and scenario 1.1.V), H.323 signaling protocol in combination with G.711 voice Codec in a homogenous network environment (i.e. voice traffic, in this study) was configured and impact of VPN (GRE tunnel, in this study) was investigated. Result analysis of the experiment showed that VPN, in this case GRE tunnel, in this set of experiment, did not show a significant impact on end-to end delay and jitter, But GRE Tunnel increased call setup time by 0.4 Seconds. Call quality (MOS) also was not affected by GRE Tunnel. In the second set of simulation experiments (i.e. Experiment.1HHE: scenario 1.2 and scenario 1.2.V), H.323, in conjunctions with G.711 in a heterogeneous network environment (i.e. data, mixed with voice traffic) was configured and impact of VPN (GRE tunnel, in this study) was examined. In this set of experiment (i.e. Experiment.2HHE), call set up time was increased slightly with GRE Tunnel, but end-to-end delay and voice quality remained unchanged (same as pervious experiment). In this set of experiment, however negative jitter was observed. Based on the findings from Experiment.1HHO and Experiment.1HHE, the end-to-end delay time in Experiment.1HE, where GRE Tunnel is established, was much greater than that of Experiment.1HHO, where no GRE Tunnel was in place. Likewise, MOS value in Experiment.1HHE was much greater than MOS value in Experiment.1HHO. Based on the comparative result analysis of Experiment.1HHE and Experiment.1HHO, the end-to-end delay and MOS value were not acceptable in a non-ideal (heterogeneous) network environment. Table 4, summarizes result analysis of impact pf GRE Tunnel in Experiment.1HHO and Experiment.1HHE.

In the next set of experiments (i.e. Experiment.2SHO: scenario 2.1 and scenario 2.1.V), in order to experiment and evaluate impact of GRE Tunnel on a SIP-based VOIP communication, we configured SIP signaling protocol and G.711 voice Codec in a homogeneous network environment (i.e. voice traffic, in this study). Result analysis of this experiment (i.e. Experiment.2SHO) showed no indication that GRE Tunnel may lead to lowering the VOIP performance in terms of end-to-end delay, call setup time, jitter and MOS. Subsequently we performed the next set of experiment (i.e. Experiment.2SHE: scenario 2.2 and scenario 2.2.V) in which SIP was configured as the choice of signaling protocol and G.711 was configured as the voice Codec, in a heterogeneous network environment, where nonVOIP traffic (i.e. web, email, database, and FTP generated along with voice traffic in the network. Based on result analysis of this experiment (Experiment.2SHE), GRE Tunnel didn’t increase QoS parameters, under investigation (i.e. end-to-end delay, call setup time, and MOS value). However, in this experiment, similar to Experiment.2HHE, we noticed a negative jitter value. In summary, performance values for Experiment.2SHE (scenario 2.2 and scenario 2.2.V) indicated that GRE Tunnel did not bring any performance issues (excluding observance of negative jitter values) in both homogeneous and heterogeneous network environments. However, in a heterogeneous (non-ideal) network environment, performance values are not acceptable. For example in scenario 2.2 and 2.2.V, the values for end-to-end delay were 5 seconds. Likewise MOS call values were 1.0 (one). According to standard voice quality measures (as outlined in tables 1 and table 3), the values for the non-ideal environment (in this study, scenarios 1.2, 1.2.V, 2.2, and 2.2.V) indicated a poor VOIP quality. Table 4, summarizes result analysis of impact pf GRE Tunnel in Experiment.2HSO and Experiment.2HSE. Table 5, provides a summary of comparative analysis on impact of GRE tunnel on different VOIP call quality measure in different scenarios (i.e. H.323 vs. SIP and homogenous vs. heterogeneous network environments).

Table 4. Result Summary- Experiment.1HHO and Experiment.1HHE

Experiment	Scenario	Signaling Protocol	Voice Codec	Traffic Type	VPN Established [Yes/No]	QOS Performance [Delay][CST*][MOS**]	Result
1HHO	1.1	H323	G.711	Homogeneous	No	Acceptable [delay: 250 ms, CST: 1.9 sec, MOS: 2.85]	VPN didn’t show a significant impact on call quality
1HHO	1.1.V	H323	G.711	Homogeneous	Yes	Acceptable [delay: 250 ms, CST: 2.3 sec, MOS: 2.8]
1HHE	1.2	H323	G.711	Heterogeneous	No	Poor [delay: 3400 ms, CST: 1.9 sec, MOS: 1.5]	VPN didn’t show a significant impact on call quality. Call QOS measures are very poor when traffic is heterogeneous
1HHE	1.2.V	H323	G.711	Heterogeneous	Yes	Poor [delay: 3400 ms, CST: 2.0 sec, MOS: 1.5]
2SHO	2.1	SIP	G.711	Homogeneous	No	Acceptable [delay: 250 ms, CST: 19 sec, MOS: 2.85]	VPN didn’t show a significant impact on call quality
2SHO	2.1.V	SIP	G.711	Homogeneous	Yes	Acceptable [delay: 250 ms, CST: 22 sec, MOS: 2.85]
2SHE	2.2	SIP	G.711	Heterogeneous	No	Poor [delay: 5000 ms, CST: 32 sec, MOS: 1]	VPN didn’t show a significant impact on call quality. Call QOS measures are very poor when traffic is heterogeneous
2SHE	2.2.V	SIP	G.711	Heterogeneous	Yes	Poor [delay: 5000 ms, CST: 32 sec, MOS: 1]
* CST: Call Setup Time * *MOS: Mean Opinion Score

In this study we also performed a comparative analysis on impact of GRE Tunnel on performance of VOIP communications, where H.323 is the signaling protocol, in a homogenous environment between G.711 and G.723 voice Codecs. Based on result of this comparative analysis, G.723 showed higher end-to-end delay and lower voice quality (in terms of MOS value). G.711, despite consuming more bandwidth, showed better performance (i.e. significantly lower end-to-end delay and higher MOS value) as compared with G723.

When we performed a comparative analysis on impact of GRE Tunnel on performance of VOIP communications, where SIP is the signaling protocol, in a homogenous environment between G.711 and G.723 voice Codecs, G.711, despite consuming more bandwidth, showed a performance equal to that of G.723. Table 6, summarizes results of comparative analysis on impact of communications. GRE Tunnel on H.323-based versus SIP-based VOIP

Table 5. H.323 vs. SIP and GRE Tunnel Impact

Experiment	Scenario	Signaling Protocol	Voice Codec	VPN Established [Yes/No]	Traffic Type	QOS Performance [Delay][CST*][MOS**]	Result
1HHO	1.1.V	H323	G.711	Yes	Homogeneous	Delay: 250 ms	VPN didn’t show a significant impact on call quality. Call QOS measures are very poor when traffic is heterogeneous
1HHE	1.2.V				Heterogeneous	Delay: 3400 ms
1HHO	1.1.V	H323	G.711	Yes	Homogeneous	CST: 2.3 sec
1HHE	1.2.V				Heterogeneous	CST: 2.0 sec
1HHO	1.1.V	H323	G.711	Yes	Homogeneous	MOS: 2.8
1HHE	1.2.V				Heterogeneous	MOS: 1.5
2SHO	2.1.V	SIP	G.711	Yes	Homogeneous	Delay: 250 ms	VPN didn’t show a significant impact on call quality. Call QOS measures are very poor when traffic is heterogeneous
2SHE	2.2.V				Heterogeneous	5000 ms
2SHO	2.1.V	SIP	G.711	Yes	Homogeneous	CST: 22 sec
2SHE	2.2.V				Heterogeneous	CST: 32 sec
2SHO	2.1.V	SIP	G.711	Yes	Homogeneous	MOS: 2.85
2SHE	2.2.V				Heterogeneous	MOS: 1

* CST: Call Setup Time

** MOS: Mean Opinion Score

Table 6. G.711 vs. G.723, on H.323-based vs. SIP-based VOIP communication

Signaling Protocol	Voice Codec	Traffic Type	VPN Established [Yes/No]	QOS Performance [Delay][CST*][MOS**]	Result
H323	G.711	Homogeneous	Yes	Acceptable [delay: 250 ms, CST: 2.3 sec, MOS: 2.8]	G.711, despite consuming more bandwidth, showed better performance (i.e. significantly lower end-to-end delay and higher MOS value) as compared with G723
H323	G.723	Homogeneous	Yes	Poor [delay: 310 ms, CST: 2.0 sec, MOS: 1.8]
SIP	G.711	Homogeneous	Yes	Acceptable [delay: 250 ms, CST: 22 sec, MOS: 2.85]	G.711, despite consuming more bandwidth, showed a performance equal to that of G.723
SIP	G.723	Homogeneous	Yes	Acceptable [delay: 250 ms, CST: 22 sec, MOS: 2.85]
* CST: Call Setup Time * *MOS: Mean Opinion Score

• V.    Conclusion and Future Works

In this paper OPNET Modeler 17.5 was used to simulate behavior of a VOIP communication running over a GRE VPN tunnel. Analysis of the result of this study indicated that VPN (in this experiment GRE Tunnel), didn’t lead to a significant increase in such quality of service (QoS) performance factors as: end-to- end delay, call setup time, or a decrease in call MOS value. However, in a non-ideal (heterogeneous) network environment, where voice and non-voice traffic coexist together, performance values for end-to-end delay, call setup time and call MOS are not acceptable according to the current standards of voice communication. No significant evidence was found to suggest that such poor results in the heterogeneous network environment can be attributed to GRE Tunnel. So, it can be concluded that heterogeneity of the network traffic, rather than use and deployment of a GRE Tunnel can be the factor responsible for lowering quality of VOIP in such network settings. We believe that deploying Quality of Service (QoS) features in a network environment can improve quality of VOIP performance. Hence, further experiments evaluating performance of VOIP communications over IP VPN in a non-ideal network environment, where QoS parameters are configured, can provide a better insight pertaining possible performance improvements in VOIP call quality in a non-ideal VPN protected network. In addition, in order to reduce the number of factors impacting performance quality of a VOIP communication, we chose to generate light volumes of voice and nonvoice (data) traffic in this experiment. In future works, role of congestion, a common issue in a non-ideal network environment, may also be examined. Therefore, experiments with different traffic distribution pattern, resembling a real life scenario may be insightful for us. Simulation of VOIP over IP VPN in a non-ideal packet switched network, combined with congestion issue and deployment of QoS parameters can provide more realistic figures relating to a real network environment.