Traffic

OBJECTIVES:
Upon completion of this chapter, the student will be able to: Define and describe the term traffic Define and describe the term Grade of Service (GoS) Use Erlangs B-table to dimension the number of channels needed in the system Explain channel utilization

Table of Contents
TRAFFIC AND CHANNEL DIMENSIONING............................................ ..01 CHANNEL UTILIZATION.............................................................................0

TRAFFIC AND CHANNEL DIMENSIONING

Cellular system capacity depends on a number of different factors. These include: The number of channels available for voice and/or data The grade of service the subscribers are encountering in the system Traffic theory attempts to obtain useful estimates, for example, the number of channels needed in a cell. These estimates depend on the selected system and the assumed or real behavior of the subscribers. What is traffic? Traffic refers to the usage of channels and is usually thought of as the holding time per time unit (or the number of call hours per hour) for one or several circuits (trunks or channels). Traffic is measured in Erlangs (E), for example, if one subscriber is continuously on the telephone, this would generate one call hour per hour or 1 E of traffic. How much traffic can one cell carry? That depends on the number of traffic channels available and the amount of congestion which is acceptable (by both the customer and the provider), the so-called Grade of Service (GoS). Different assumptions on subscriber behavior lead to different answers to this question. Erlangs (a Danish traffic theorist) Btable is based on the most common assumptions used. These assumptions are: No queues Number of subscribers much higher than number of traffic channels available No dedicated (reserved) traffic channels Poisson distributed (random) traffic Blocked calls abandon the call attempt immediately This is referred to as a loss system. Erlangs B-table relates to the number of traffic channels, the GoS, and the traffic offered. This relationship is tabulated in Figure 4-1. Assuming that one cell has two carriers, corresponding typically to 2x8-2=14 traffic channels and a GoS of 2% is acceptable, the traffic that can be offered is A=8.2003 E (Figure 4-1). This number is interesting, if an estimate of the average traffic per subscriber can be obtained. Studies show that the average traffic per subscriber during the busy hour is typically 15-20mE (this can correspond to, for example, one call lasting 54-72 seconds per hour). Dividing the traffic that one cell can offer, Acell=8.20 E, by the traffic per subscriber, here chosen as Asub=0.025 E, the number of subscribers one cell can support is derived as

8.20/0.025 = 328subscribers.
n .007 . 008
.009 .01 .02 .03 .05 .1 .2 .4 n 1 .00705 .00806 .00908 .01010 .02041 .03093 .05263 .11111 .25000 .66667 1 2 .12600 .13532 .14416 .15259 .22347 .28155 .38132 .59543 1.0000 2.0000 2 3 .39664 .41757 .43711 .45549 .60221 .71513 .89940 1.2708 1.9299 3.4798 3 4 .77729 .81029 .84085 .86942 1.0923 1.2589 1.5246 2.0454 2.9452 5.0210 4 5 1.2362 1.2810 1.3223 1.3608 1.6571 1.8752 2.2185 2.8811 4.0104 6.5955 5 6 1.7531 1.8093 1.8610 1.9090 2.2759 2.5431 2.9603 3.7584 5.1086 8.1907 6 7 2.3149 2.3820 2.4437 2.5009 2.9354 3.2497 3.7378 4.6662 6.2302 9.7998 7 8 2.9125 2.9902 3.0615 3.1276 3.6271 3.9865 4.5430 5.5971 7.3692 11.419 8 9 3.5395 3.6274 3.7080 3.7825 4.3447 4.7479 5.3702 6.5464 8.5217 13.045 9 10 4.1911 4.2889 4.3784 4.4612 5.0840 5.5294 6.2157 7.5106 9.6850 14.677 10 11 4.8637 4.9709 5.0691 5.1599 5.8415 6.3280 7.0764 8.4871 10.857 16.314 11 12 5.5543 5.6708 5.7774 5.8760 6.6147 7.1410 7.9501 9.4740 12.036 17.954 12 13 6.2607 6.3863 6.5011 6.6072 7.4015 7.9667 8.8349 10.470 13.222 19.598 13 14 6.9811 7.1154 7.2382 7.3517 8.2003 8.8035 9.7295 11.473 14.413 21.243 14 15 7.7139 7.8568 7.9874 8.1080 9.0096 9.6500 10.633 12.484 15.608 22.891 15 16 8.4579 8.6092 8.7474 8.8750 9.8284 10.505 11.544 13.500 16.807 24.541 16 17 9.2119 9.3714 9.6171 9.6516 10.656 11.368 12.461 14.522 18.010 26.192 17 18 9.9751 10.143 10.296 10.437 11.491 12.238 13.385 15.548 19.216 27.844 18 19 10.747 10.922 11.082 11.230 12.333 13.115 14.315 16.579 20.424 29.498 19 20 11.526 11.709 11.876 12.031 13.182 13.997 15.249 17.613 21.635 31.152 20 21 12.312 12.503 12.677 12.838 14.036 14.885 16.189 18.651 22.848 32.808 21 22 13.105 13.303 13.484 13.651 14.896 15.778 17.132 19.692 24.064 34.464 22 23 13.904 14.110 14.297 14.470 15.761 16.675 18.080 20.737 25.281 36.121 23 24 14.709 14.922 15.116 15.295 16.631 17.577 19.031 21.784 26.499 37.779 24 25 15.519 15.739 15.939 16.125 17.505 18.483 19.985 22.833 27.720 39.437 25 26 16.334 16.561 16.768 16.959 18.383 19.392 20.943 23.885 28.941 41.096 26 27 17.153 17.387 17.601 17.797 19.265 20.305 21.904 24.939 30.164 42.755 27 28 17.977 18.218 18.438 18.640 20.150 21.221 22.867 25.995 31.388 44.414 28 29 18.805 19.053 19.279 19.487 21.039 22.140 23.833 27.053 32.614 46.074 29 30 19.637 19.891 20.123 20.337 21.932 23.062 24.802 28.113 33.840 47.735 30 31 20.473 20.734 20.972 21.191 22.827 23.987 25.773 29.174 35.067 49.395 31 32 21.312 21.580 21.823 22.048 23.725 24.914 26.746 30.237 36.295 51.056 32

Figure 4-1 Part of Erlangs B-table, yielding the traffic (in Erlangs) as a function of the GoS (columns) and number of traffic channels (rows) Dimensioning the network now implies using demographic data to determine the sizes of the cells. The preceding example is simplified, however, it provides an understanding of what is meant by traffic and traffic dimensioning.

The problem may be that given a number of subscribers in one particular area (for example, an airport), how many carriers do we need to support the traffic, if only one cell is to be used? Dimensioning a whole network while maintaining a fixed cell size means estimating the number of carriers needed in each cell. In addition, traffic is not constant. It varies between day and night, different days, and with a number of other factors. Mobile telephony implies mobility and hence subscribers may move from one area to another during the course of a day. It is important that the number of signaling channels (SDCCHs) is dimensioned as well, taking into account the estimated system behavior in various parts of the network. For example, cells bordering a different location area may have lots of location updating, and cells on a highway probably have many handovers. In order to calculate the need for SDCCHs, the number of attempts for every procedure that uses the SDCCH, as well as the

time that each procedure holds, the SDCCH must be taken into account. The procedures are; location updating, periodic registration, IMSI attach/detach, call setup, SMS, facsimile, and supplementary services. The number of false accesses must also be estimated. This is typically quite a high number, but still small, compared to the traffic. If A is the traffic on the SDCCH for normal call and fax setup and A' is the traffic that accounts for the rest of the procedures performed on the SDCCH, we obtain the global grade of service, GoST, for calls which go through an SDCCH and a TCH: GoST = GoS1 + (1-GoS1)GoS2 where GoS1 is the grade of service on the SDCCH and GoS2 is the grade of service on the TCH. Since we can assume that the grade of service given by the operator corresponds to the traffic channels, thatis, GoS2, it is obvious that a better grade of service is needed on the SDCCH. This is not only important because signaling is performed on it, but also because the global grade of service would otherwise be higher than the grade of service expected by the operator. The most accurate SDCCH dimensioning is achieved by looking at the congestion level in the TCHs and the SDCCHs for the specific cell. The optimum configuration is achieved by selecting a configuration with as many TCHs as possible, without letting the GoS1 exceed 1/4 of GoS2 (with only four SDCCHs, GoS1 GoS2/2). SDCCHs can only be allocated in steps of four or eight, as required by the GSM specifications, that is, one cell can have four, eight, twelve, etc., SDCCHs, but no more than 128 SDCCHs can be allocated to one cell.

CHANNEL UTILIZATION
Assume the task is to find the necessary number of traffic channels for one cell to serve subscribers with a traffic of 33 E. The GoS during the busy hour is not to exceed 2%. By considering the above requirements and consulting Erlangs B-table, 43 channels are found to be needed (Figure 4-3).
n .007
008 .009 .01 .02 .03 .05 .1 .2 .4 n 43 30.734 31.069 31.374 31.656 33.758 35.253 37.565 42.011 49.851 69.342 43

Figure 4-3 Part of Erlangs B-table for 43 channels giving the offered traffic (E) as a function of the GoS (%)

Assume five cells are designed to cover the same area as the single cell. These five cells must handle the same amount of traffic as the cell above, 33 E. Acceptable GoS is still 2%. First, the total traffic is divided among the cells (Figure 4-4). Traffic distribution over several cells results in a need for more channels than if all traffic had been concentrated in one cell.

This illustrates the fact that it is more efficient to use many channels in a larger cell than vice versa. To calculate the channel utilization, the traffic offered is reduced by the GoS of 2% (yielding the traffic served) and dividing that value by the number of channels (yielding the channel utilization). With 43 channels (as in the previous single cell example), the channel utilization is 33.083/ 43 = 77%, that is, each channel is used approximately 77% of the time. However, by splitting this cell into smaller cells, more traffic channels are required, hence, the channel utilization decreases.
Cell A B C D E Traffic (%) 40 25 15 10 10 100 Traffic (E) 13.20 8.25 4.95 3.30 3.30 33.00 No. of channels 21 15 10 8 8 62 Channel utilization (%) 62 54 49 40 40

Figure 4-4 What happens when a certain amount of traffic is distributed over several cells?

As we will see in the following chapter, capacity and interference problems prevent us from always using the most effective channel utilization scheme and, thus, solutions in real networks must compromise between efficiency (cost) and quality.