IP Telephony development and performance over IEEE g WLAN

Transcription

1 IP Telephony development and performance over IEEE 8.11g WLAN Miguel Edo 1, Miguel Garcia, Carlos Turro 3 and Jaime Lloret Universidad Politécnica de Valencia, Camino Vera s/n,, Valencia (Spain) 1 miedmon@epsg.upv.es; migarpi@posgrado.upv.es; 3 turro@cc.upv.es; jlloret@dcom.upv.es Abstract With the adoption of the Wireless LAN technology as one of the main ways to access the enterprise network, IP services have found another place where they can be implemented. In this paper we will show the test and performance used to develop the IP telephony network over the 8.11g wireless LAN of the Polytechnic University of Valencia. In order to make these measurements, we have used the Open Source PBX & Telephony Platform: Asterisk and SmartPhones. We will show the results obtained about the delay, the jitter and the number of lost packets when the SmartPhones are in the same wireless cell and when they are roaming. Finally, we will calculate the amount of IP phones that can be working in a single access point and the bandwidth wasted in different cases. This work can be used to design VoIP and IP telephony wireless networks and to design wireless IP phones relocation algorithms. 1. Introduction IP telephony, or VoIP (Voice over IP), enables voice communications over networks based on the Internet Protocol (IP). Therefore, it allows significant advantages. On one hand, IP phones communications within the intranet are free. This is most interesting for companies or institutions which have several branches or for mobile employers who are moving inside the intranet with their mobile devices. On the other hand, the huge investments that have to be done by the companies or institutions to purchase a Private Branch Exchange (PBX) can be reduced by using PBXs based on free software. They provide the same functionality as a traditional PBX. From several years ago, wireless networks have been achieving great popularity because the deployment of these networks are low cost while provide us quite mobility and scalability [1]. These networks have evolved quickly to meet the needs of these users: more security and bandwidth. It is therefore evolved from IEEE8.11b, which is used to supply a theoretical bandwidth of 11Mbps, to IEEE8.11a and IEEE8.11g [] which provides a theoretical bandwidth of Mbps. This technology is always in progress. Although IP telephony was firstly deployed for the wired network, it can also be deployed on the wireless network. One of the main advantages of the IP telephony over a wireless network is that it allows mobility of the people while they are talking. Currently, most PDA's and SmartPhones incorporate Wireless LANs connectivity that is a considerable advantage because such devices can be used as either cell or VoIP phone. It is a very important feature because as long as we have WLAN coverage we will be able to make VoIP calls. In this paper we will show how an IEEE 8.11g WLAN performs using IP telephony with a PBX: Asterisk [3]. The rest of the paper is as follows. Section gives previous studies and implementations of IP Telephony. Section 3 shows the network deployment and the main features of the IP PBX. In Section, we will show the measurements of the delay, jitter, packet loss and bandwidth for different scenarios. Finally, section will present the conclusions.. Related Work In the literature, we can find several publications dealing with IP telephony over wireless networks. In some sections there is a discussion about the ability of the IEEE 8.11b/g networks to handle VoIP traffic. In such papers there are theoretical studies on the feasibility of IP telephony over WLAN [] [] and also, in some of them, the quality of calls using different audio codecs G.711 and G.719 [] is checked. In other papers, the behavior of IP telephony over WLAN, when these wireless networks have another kind of traffic, is analyzed []. Furthermore, some studies have been conducted on IP Telephony Asterisk-based PBX [7], where an

2 experimental assessment of the IEEE 8.11b standard to support VoIP on a wired network has been carried out. Nevertheless, none of the studies aforementioned has dealt with IEEE 8.11g networks. Neither any of them have studied how roaming affect the phones or the required bandwidth for the phones to have enough quality of service. Eventually, none of these considerations have been applied to the SmartPhones case. UPV Network AP Cisco Aironet 113AG IP PBX Asterisk 3. Network Deployment and IP PBX The WLAN of the Polytechnic University of Valencia [8] is formed by 7 access points (APs) spread in 3 campuses. 33 of them are in the Campus of Gandia, APs are in the Campus of Alcoy and APs are in the main Campus (Campus de Vera). The access points are Cisco Aironet 113AG Series APs which use IEEE 8.11a/b/g standard and provide speeds up to 18Mbps. They are installed to allow the users a continuous coverage as they roam throughout a facility. They incorporate the 8.11i IEEE-Compliant standard (WPA-Certified and WPA-Certified) which allows interoperability with other manufacturers. The coverage of each access point varies between 3 m at Mbps and 137 m at 1 Mbps for indoor environments. The IP Telephony PBX is a server that runs a software desktop application. Asterisk is a freesoftware application (under GPL) performing a function as a regular telephone PBX. Such as many PBX, it is possible to connect a specific amount of phones to make calls between each other and even to connect to a VoIP provider or to the PSTN (Public Switched Telephone Network). The basic package of Asterisk includes many features that were previously available only in expensive proprietary systems such as creation of extensions, sending voice messages to s, conference calls, voice interactive menus and automatic call distribution. Asterisk supports various VoIP protocols including SIP (Session Initiation Protocol) [9]. SIP is a protocol for controlling and signaling systems used primarily in IP Telephony which was developed by the IETF (RFC 31). The protocol allows to start, to modify and to finalize multimedia sessions with one or more participants and its greatest advantage lies in its both simplicity and consistency. The audio codecs supported be the Asterisk are the following ones: G.711 ulaw, alaw G.711, G.73.1, G.7, G.79, GSM, ilbc, LPC1, Speex. Figure 1. Network architecture.. Real Measurements In this section we will show the measurements carried out in our experiment to evaluate the network performance..1. Test Bench IP Phone Nokia E In order to test the network performance and analyze which features offers, we will use two SmartPhones Nokia E and an Asterisk VoIP telephone PBX. The SmartPhones will be connected to the Asterisk PBX through the wireless network of the Polytechnic University of Valencia. The IP telephones incorporate the IEEE 8.11g standard and WPA encryption. This is necessary because the connection to the wireless network of the Polytechnic University of Valencia is established by using IEEE 8.11g, WPA encryption with Protected EAP (PEAP) and EAP-MSCHAP v authentication, thus ensuring a secure communication. These phones support the SIP protocol which we use to connect with the Asterisk PBX. The packages that have been captured for further study are RTP packets on UDP/IP using the network analyzer Wireshark [1]. As for the audio codec, we will use the G.711. This codec will give us the best voice quality as it does not use any compression. It is the same codec used by the network ISDN (Integrated Service Digital Network) and the sound quality is like a conventional telephone. It also has the lowest latency since there is no need for compression, which leads to less processing load.

3 Delay (ms) Call Fig. Measures of the call delay Delay (ms) Delay average (ms) Max Delay (ms) Call Call Call Call Call8 Fig 3. Delay average and maximum delay per call. Call Jitter average (ms) Max Jitter (ms) Jitter (ms) 3 1 Jitter (ms) Fig. Measures of the call jitter Call Call Call Call Call8 Fig. Jitter average and maximum jitter per call.. 3 second call In order to analyze the performance and quality of calls we have made 9 calls of 3 second each one, testing the delay, jitter, bandwidth and packet loss. The packets are captured from the IP Phone to the Asterisk PBX...1 Delay tests The Figure shows the data obtained in the delay measurements. It only shows out of 9 calls because we want the graph to be understandable. As we can see, the calls are around 3 second long and none of them exceed ms delay. In Figure 3 we have the highest average delay and delay per call. Visually we can see that the average delay time is around ms. Performing the overall average of 9 calls we obtain a delay of ms and the average maximum delay is 39.1 ms.. Jitter testing In Figure we can see the results of tests done about Jitter. In neither of the cases, the jitter exceeds ms and is maintained almost constant all the time around 1 ms. In Figure we have an average of the jitter and maximum jitter per call. In the graph we see that the jitter in none of the 9 calls exceeds ms giving an average of ms jitter and 3.9 ms of maximum jitter average...3 Lost packets testing In Figure we can see the results of the tests carried out concerning lost packets. In calls 1 and we can see how the number of lost packets grows rapidly around 1 seconds after the call starts because the IP phone buffer is filled up. Nevertheless, under no circumstances is this packet loss appreciable during the communication. In the following figures 7 and 8 we can see that the packet loss in calls is very high. The packet loss average is 1. packages: a 1.39 % packet lost. Although it may seem a very high rate, this does not affect the conversation. The 9 calls have been of excellent quality. The transmitted packets average is for the 9 3-second calls.

4 Lost Packets Call Fig.. Extent the packets lost in the calls. Lost Packets 1 Lost packets Packets/Call average Call Call Call Call Call8 Fig 7. Transmitted packets and lost packets per call.,%,% Lost Packets average (%) Call Lost Packets 1,% 1,% Delay (ms) 3,% 1,% Call Call Call Call Call8 Fig 8. Percentage of lost packets per call... Roaming testing It has done the same procedure as in section. but in this case we have made two calls. One of the phones is static in an Access Point and the other is moving around the wireless network at the UPV in the Campus of Gandia. The data displayed on the following points are those obtained from the phone on the move...1 Testing Delay The figure 9 shows the data obtained in the analysis of the delay in testing Roaming. As we can see the same thing that happens in figure 1, the delay is kept around ms but in this case because the information has to go through many more network equipment ranges between ms and 1 ms with an average of ms very similar to data obtained in section..1. In this test, the maximum average delay is 9.9 ms whereas it was 39.1 ms in the 3-second calls between the two phones using the same Access Point... Jitter Testing In figure 1 we can see the data obtained in the analysis of jitter in the roaming test. As shown in the figure, jitter is around. and 1. ms very similar to tests carried out in section.. with the difference in Time (ms) Fig 9. Extent of the delay in Roaming this case that the maximum jitter average of 3.9 ms increases to.3ms. On the other hand, the average jitter has not increased over the average obtained in section..: 1. ms this time, which is a positive development...3 Lost packets test In figure 11 we can see the data obtained in the analysis of roaming calls regarding lost packets. As in previous points, we have something similar to what happened in the 3-second calls. In this case we have a lost packet average per call of 8. % without causing any problem in this communication. On the other hand, the maximum packet loss is caused at times when the buffer is full in the IP phones. As we can see, these maximum losses are equidistant... Test of effective bandwidth In figure 1 we show the test of effective bandwidth in the network of the Polytechnic University of Valencia. These tests show us that we have a capacity of around Kbps with an average of Kbps in the IEEE 8.11g wireless network. Then, in the following paragraphs we will see the bandwidth occupied by IP phones and then will calculate the amount of IP phones that theoretically might work in this wireless network.

5 7 Call 1 1 Call 1 Jitter (ms) 3 Lost Packets Fig 1. Measures on Roaming Jitter Time (ms) Fig. 11. Lost packets in roaming 1 BW (Kbps) Fig 1: Effective bandwidth in the IEEE 8.11g network. IP BW (Kbps) 8 Call Fig. 13. Measures of the 3-second call bandwidth..1 Bandwidth test using the G.711 audio codec At this point we are going to see the occupied bandwidth both by the set of the 9 of 3-second calls and by the roaming calls, using in both tests the audio codec G.711 In Figure 13, we can see how the calls 1 and have a drop of bandwidth, this is due to packet loss because the buffer of the IP phone in both cases is full. This was explained earlier in paragraph..3. On the other side, in the moments where there is no loss of packets we can see that the bandwidth is around 8 and 11 Kbps with an average of 89Kbps. In figure 1 we can see the average bandwidth per call that is always between 1 and 1Kbps: concretely 11. Kbps. This is very important because then we will calculate the theoretical number of phones that can operate on the wireless network of the UPV. In Figure 1, we may see something similar that happens with the 3-second call bandwidth. On this occasion, the bandwidth has a mean of 9.19 Kbps very similar to 89Kbps in the 3 second s average calls. On the other hand, what does vary significantly is the maximum bandwidth average which raises from 11. Kbps up to 18Kbps. This is very important in order to calculate the maximum number of phones. According to the paragraph above the effective bandwidth in the IEEE 8.11g network of the Polytechnic University of Valencia is Kbps. In our case, the maximum bandwidth that generates our phones when they were in a position to roaming (worst situation) was 18Kbps. Following these steps we can say that the theoretical number of phones that our wireless network support per access point is approximately 1.. Conclusions In conclusion, we can say that the IEEE 8.11g wireless network of the Polytechnic University of Valencia could, theoretically, support up to 1 IP phones per access point using the audio G.711 codec. This number would be obtained in an ideal situation where we had always this effective bandwidth with a small amount of external interference always in a network devoted solely to IP telephony without any other type of traffic. This can be improved thanks to the system proposed by the same authors of this paper [11].

6 1 IP BW average (kbps) Max IP BW (kbps) IP BW (Kbps) 1 8 IP BW (Kbps) 1 8 Call Call Call Call Call Call8 Fig. 1. Measure of the maximum bandwidth average and maximum bandwidth per call. 3 Call Roaming Call Delay 19.7 ms ms Jitter 1.1 ms 1. ms IP BW 89 Kbps 9.1 Kbps Lost Packet % 1.39% 8. % Table 1. Mean value of the data obtained in the tests aforementioned. On the one hand, we have realized that the average of the data obtained in both cases in discussion (the set of 9 3-second calls and the roaming calls) are very similar. This can be seen in Table 1. By contrast, in table we can see as the maximum increases markedly in the roaming calls. This is due to the mobility of the user who makes a reassociation between different access points constantly necessary. We also consider it could be interesting to use SmartPhones with IP phones and WiFi connectivity. These phones have the ability to connect to a wireless network, and then use the same device to make VoIP and cellular calls. Our future work will be focused on carrying out performance tests connecting the Asterisk PBX with a standard PBX in order to make calls outside the Polytechnic University of Valencia. We will make those calls using the PSTN (Public Switched Telephone Network) and with another Asterisk s supported audio codecs, e.g. G.73.1, G.7 and G.79. It will open new research lines about mixed standard mobile-wireless IP Telephony architectures.. References [1] Kit-Sang Tang, Kim Man-Fung and S. Kwong, Wireless Communication Network in IC Design Factory, IEEE. Transactions on industrial electronics, vol. 8, No., pp. -9, Hong Kong, April Fig. 1. Roaming bandwidth measures 3s Call Roaming Call Delay 39.1 ms 9.9 ms Jitter 3.9 ms.3 ms IP BW 11. Kbps 18 Kbps Table. Average of the maximum values. [] IEEE It is available in The Working Group for WLAN Standards [3] Asterisk. It is available enwww.asterisk.org [] D.P. Hole, F.A. Tobagi, "Capacity of an IEEE 8.11b wireless LAN supporting VoIP" IEEE International Conference on Communications, vol.1, pp. 19-1, -, Paris (France), June. [] L. Cai, Y. Xiao, X. Shen and J. W. Mark, "Voice Over IP- Theory and Practice." International Journal of Communication Systems, vol. 19, Issue, pp April. [] A. Dutta, P. Agrawal, S. Das, et al. "Realizing mobile wireless Internet telephony and streaming multimedia testbed," Computer Communications, vol. 7, Issue 8, May, Pages [7] G. Agreda, J. Gaviria. "EvaluaciónExperimental the Capacity of IEEE 8.11b support for VoIP." Converging technologies applied to mobile computing. Memories IComM. Pp [8] Jaime Lloret Mauri, Jose Javier López Monfort and German Ramos, Wireless LAN Deployment Extension in Large Areas: The Case of a University Campus, Communication Systems and Networks 3, Benalmadena, Malaga (Spain), September 3. [9] RFC 31, SIP: Session Initiation Protocol. [1] Wireshark 1... It is available in [11] Miguel Garcia, Diana Bri, Carlos Turró, Jaime Lloret. A User-Balanced System for IP Telephony in WLANs. The Second International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies, 8. UBICOMM'8. Publication Date: Sept. 9 8-Oct. 8 On page (s): 1-.