IP telephony network saving capacity due to substitution of PSTN by IP network
Milutin Kapov1, Damir Dlaka2
1FESB-Split, University of Split
2 Ericsson Nikola Tesla, Split
Abstract: IP telephony has many advantages over traditional PSTN telephony. These advantages are mostly manifested in the possibility of implementing many new services as well as in the network capacity saving. The focus of this paper is the IP network transfer capacity savings analysis. IP network completely replaces the PSTN network. Also, the influence of the voice activity detection technique and VoIP packet header compression technique on the additional saving of the required IP network transportation capacity is analyzed. The analysis is performed on the basis of the simulation measurements results which have been obtained by the originally developed application PlanVoip. The obtained results of the simulation measurements are presented in the analytical and graphic form.
1. INTRODUCTION
Many possibilities are opening with the development of new technologies. The same thing is with telephony. Today it is possible to change PSTN telephony based on the circuit switching with the IP network based on the packet switching. When the Internet or other IP network is used for the voice transfer, we talk about IP telephony. The key moment for developing and accepting IP telephony has happened with gateway introduction, which has enabled the interconnection between PSTN and Internet/IP world. Also, the gateway supports traditional PSTN telephones in the IP telephony world, which is very important for the ordinary user [1], [2].
There are many advantages of the IP telephony. The great advantage of the IP telephony over PSTN is the number of new services which cannot be supported with standard telephony. Also, IP telephony requires less transmission capacity which results in higher efficiency of the existing transmission resources.
Besides capacity saving which is obtained because of the IP network principles (packet switching), there are techniques which additionally contribute to capacity savings. These techniques are voice activity detection and VoIP packet header compression.
The aim of this article is capacity saving analysis which is done by substituting the PSTN network with the IP network. The analysis is based on the expected traffic which defines the resources needed for the PSTN network. For the same traffic assumption there are parameters defined for IP network. Based on the comparison analysis of the both networks, IP network advantages over PSTN in the transmission capacity domain become evident.
Complex comparative analysis of the different values of the relevant parameters influence with using originally developed integral application PlanVoip is done. Until now it is not existed, as we know, integral application which embraces all relevant parameters, and achieves quick and mathematical analysis of these parameters influence on the PSTN and, especially on the IP traffic. Efficiency of PlanVoip application can be seen in quick and mathematical correct defining of the planed network resources based on the elected relevant parameters values. With that it is achieved significant savings of the time and money, which is advantage. A detailed description of the PlanVoip application is given in [7].
In the second chapter there is a short overview of the relevant traffic parameters for the voice transfer over PSTN network. In the third chapter, important parameters which influence needed IP network transmission capacity are analyzed. In the fourth, central, chapter the calculation of the required transmission capacity for the PSTN and IP network for the same traffic load is presented. Based on the achieved results which are presented in the analytical and graphic form, comparative carried out analysis is. The last, fifth chapter contains our conclusions.
2. PSTN TELEPHONY
In this chapter, relevant traffic parameters which influence on the PSTN network dimensioning are described. Observations are mentioned so that achieves easy later analysis of the voice transfer over IP network.
Data and voice networks are defined by a number of different factors, but the most important are quality of services and the price. Balancing these two factors we want the user to be satisfied with the offered service and we want the provider to achieve profit. Correctly designed network assures satisfactory service with the low blocking level and high connection efficiency.
The basic precondition for the successful design of voice networks is correct planning of the network capacity. The required capacity is defined according to the traffic load which the network can support. Because of that, high importance is given to traffic load analysis. In that analysis it is necessary to choose a correct traffic model. Traffic models make possible the close overview of the real traffic. There are many traffic models. Which traffic model will be chosen depends of many factors.
Traffic load is defined as a amount of traffic which travels over connection in the specific time interval. There are many ways to measure traffic load. To define the required traffic capacity we analyzed used the traffic load at the most frequent hour. The average time of one call in a certain time interval AHT is defined as:
(1)
Cd total time of the all calls in a certain interval,
Nc number of calls in that time interval.
Traffic load Tb in erlangs is defined as:
(2)
The traffic load of the most frequent hour Tm defines the maximum traffic load which the network can support. Generally, it is assumed that the traffic load of the most frequent hour Tm is between 15 and 20% of the total daily load. We assumed that it is 17%, and so we have:
(3)
The described calculation takes in to consideration only the calls which are realized.
Traffic models are used in order to simulate real traffic. Although none of traffic models can accurately approximate real traffic, thery are chosen in such manner that they overlap as much as possible with real traffic. There are a few factors which influence the right choice of the traffic model. The main factors which influence the traffic model selection are [3]:
- type of incoming traffic (smooth, peak, random);
- number of blocked calls (LCH-Lost Calls Held, LCC-Lost Calls Cleared, LCD-Lost Calls Delayed, LCR-Lost Calls Retried);
- number of traffic sources (finite or infinite);
- holding time (exponential).
Taking in to consideration all the above mentioned factors, many traffic models have been developed. The most used are: Poisson, Erlang B [12] [13], extended Erlang B, Erlang C, Engset, EART/EARC and Neal-Wilkerson.
3. IP TELEPHONY
Internet has been developed in order to achieve easy voice, data and video interconnection through personal computer/notebooks. In time, preconditions have been accomplished to that the Internet could be used for telephone conversation and also for advanced communication such as video conferencing.
IP telephony makes possible all kinds of communication combinations between standard telephone, IP telephone and PC telephone. If we use the software IP telephone (PC telephone) or IP telephone in combination with standard telephones, it is necessary to use the gateway and the gatekeeper. They provide the successful interconnection of traditional telephone world with Internet/IP world [4], [5].
For the voice transfer in the IP network the transport protocol in real time is used as a solution to avoid possible stopping in communication due to loss or error inside the packet. In the analysis of PSTN substitution with IP, it is necessary to define these parameters which have influence on bandwidth defining. These parameters have a great influence on defining the required bandwidth [3], [6], [7], [8], [14]:
1. voice codec algorithm,
2. VoIP packet size,
3. number of VoIP packet per second,
4. voice activity detection (VAD),
5. RTP/UDP/IP header compression (cRTP),
6. Communication on the IP network.
1. Voice codec algorithm defines selection between many types of voice codec which are used in the IP telephony. Voice codec types are different according to coding speed and specific codec bandwidth. Codec bandwidth Bc is the number of bits per second which are necessary to generate in order to satisfy the defined bandwidth. Codec influences the bandwidth because it defines the size of useful information inside packets which are transferred across the IP part of the network. The most used codecs are: G.711, G.726, G728, G.729, G.723.1 and others.
2. VoIP packet size depends on the size and number of voice blocks which are inside them. The size of voice block is defined by the chosen codec standard. The total number of voice blocks inside one VoIP packet influences the number of packets sent in one second. These also influence the required bandwidth. Fewer voice blocks inside one VoIP packet means less useful information in a packet. This requires a higher number of VoIP packets per second resulting in a higher bandwidth. On the other hand, higher number of voice blocks inside VoIP packet reduces the required needed bandwidth, but increases the total delay for every call.
The size of one voice block Fs is a number of bytes defined by a DSP processor in every interval of voice block codec. The duration of one voice block Kb is time interval of voice block in which codec takes voice blocks. The total size of voice information inside one VoIP packet Vb and duration of voice information Vt are given in the following expressions (4):
(4)
n number of voice blocks inside a VoIP packet.
3. The number of VoIP packet per second Ps is a number of VoIP packets which need to be sent every second to satisfy the defined codec speed specified by the standard of the specific algorithm. The following expression defines Ps:
(5)
The total size of VoIP packet Sp is:
(6)
L2 header size of the frame from the 2. OSI model,
Hd RTP/UDP/IP header size (40 bytes).
One VoIP channel bandwidth Vk is given in the following expression:
(7)
4. Voice activity detection (VAD) is introduced due to the fact that voice conversation includes from 35 to 50% of silence. VAD can eliminate packets which include silence and consequently reduce the required bandwidth. The reduced bandwidth in VoIP networks with activated VAD technique is defined as:
(8)
5. RTP/UDP/IP header compression (cRTP) provides compression of the 40 bytes header on the 2, 4 or 5 bytes, with which a significant reduction of the require bandwidth for VoIP channel is accomplished. cRTP header compression technique is based on the fact that only a few fields in the header change with every packet. During the compression process changing parameters are saved in every packet. cRTP technique is performed for every subsequent of switching, but not for the whole transportation way in advance. Therefore, the compression algorithm can rebuild the original header without any loss of information.
6. Communication on the IP network is based on the packet switching. Packets can be distributed through different ways and on the receiving end are reassembled in correct order. Grouping is one of basic factors which influence the required bandwidths on the connections which are left after the breakdown on specific links. Connections which are left after the breakdown on specific links we called failover links. After the breakdown, failover links must support traffic which traveled across broken link(s). This means that, due to increased traffic, other links must be designed for additional loads [9], [10], [11].
4. SIMULATION MEASUREMENTS AND RESULT ANALYSIS
In this chapter, simulation measurements results are presented and the comparative analysis of achieved results. Simulation measurements are based on the expected traffic load in the PSTN networks. According to the expectations, the defined required capacity for the PSTN network and also is defined as well as the capacity of the IP network which must support the same traffic load. Fig. 1 shows the example of the analyzed network on which we perform the required simulation measurements of the traffic and the comparative analysis of achieved results.
Figure 1. Analyzed network
The considered network consists of headquarter and their three branches. On Fig. 1 traffic loads and relevant parameters values are defined. Simulation measurements are performed by means of developed application PlanVoip [7].
4.1. Required bandwidth in the PSTN network
The traffic which must be supported by the analyzed PSTN network has the following features:
- incoming traffic type is random,
- number of blocked calls is LCC (Lost Calls Cleared),
- number of callers is infinite,
- holding time is exponential.
According to these predictions of the incoming traffic it is best to use Erlang B traffic model. The assumed blocking calls probability on every transportation link during connecting is 5% (Gs=0.05).
We have assumed that the total time of PSTN calls duration for one day for all four locations are (Fig. 1):
headquarter Cd = 10 300 minutes/day,
branch 1 Cd1 = 4 400 minutes/day,
branch 2 Cd2 = 3 650 minutes/day,
branch 3 Cd3 = 2 300 minutes/day.
For the traffic load of the most frequent hour Tm we have taken 17% with respect to the total daily traffic load, and consequently the load of the most frequent hour for every location is:
Tm (headquarter) =[0.17 × (10 300 min/day)] ¤ 60=
=29.18 erlang
Tm (branch 1) = [0.17 × (4 400 min/day)] ¤ 60 =
=12.47 erlang
Tm (branch 2) = [0.17 × (3 650 min/day)] ¤ 60 =
=10.34 erlang
Tm (branch 3) =[ 0.17 × (2 300 min/day)] ¤ 60 =
=6.52 erlang
With Tm, Cd and Gs values for every location, and with PlanVoip application which have erlang B tables integrated into the selected traffic model, we can calculate the number of required connections Nk (DS0=64kb/s) for every location [3], [7]:
Nk (headquarter) = 35
Nk (branch 1) = 18
Nk (branch 2) = 15