Διαστασιοποίηση Δικτύου (Network dimensioning) Αν. Καθηγητής Γεώργιος Ευθύμογλου Module Title

Εισαγωγή Network Dimensioning Capacity of Base Station (BS) Coverage dimensioning Services dimensioning Capacity dimensioning Capacity of Base Station (BS) Link Budget Analysis Estimation of number of subscribers per BS Module Title

Cellular system The coverage area is usually divided in hexagons The area of a hexagon with cell radius R is given by The frequency reuse factor in the cellular system shown here is K=7. Adjacent cells use different spectrum in order to reduce other (inter) cell interference.

Cellular System: cell range The cell radius R depends on the terrain of operation. Therefore, the number of required Base Stations required to cover a service area depends on the selected cell range and the type of customer premise equipment (CPE).

Cellular system: base station capacity Given the cell range, R, the BS capacity for omnidirectional antenna can be determined in two different ways: Using the distances that each MCS is used: MCS zones, OR Using the notion of oversubscription ratio (OSR)

Spectral efficiency of MCS Assume BW=1 Hz ID MCS Spectral efficiency of MCS (bit/sec/Hz) SNR (dB) 1 BPSK 1/2 1 x ½ = 0.5 6.4 2 QPSK 1/2 2 x ½ =1.0 9.4 3 QPSK 3/4 2 x ¾ =1.5 11.2 4 16-QAM ½ 4 x ½ = 2.0 16.4 5 16-QAM 3/4 4 x 3/4 = 3.0 18.2 6 64-QAM 2/3 6 x 2/3 = 4.0 22.7 7 64-QAM 3/4 6 x ¾ = 4.5 24.4

Spectral efficiency of MCS with OFDM Assume BW = 3.5 MHz Assume OFDM efficiency = 0.8

Base station capacity Assuming N = 7

Cellular system with MCS Given the cell range the BS capacity (supported throughput) can be determined as follows: The coverage area Aj for the j-th MCS, assuming hexagonal cells is The supported BS throughput taking into account adaptive MCS zones is given by

Cellular system with MCS: Oversubscription ratio Over-subscription ratio (OSR) is the ratio of the total subscriber’s throughput demand over the reference capacity of the base station. The reference capacity of the base station corresponds to the available bit rate of the lowest modulation scheme served with that BS, termed as Cref. If at the cell edge, the MCS is QPSK ½ , the corresponding throughput will be the reference capacity. An OSR value, for example OSR=10, means that because higher throughput than the reference is achieved at distances closer to the BS (using higher order MCS, such as16-QAM), the capacity offered by the BS is much higher than Cref

Cellular system with MCS: Oversubscription ratio For example, if a WiMAX system serves 80% residential users with data rate 512Kbps and 20% business users with data rate 1Mbps, The total capacity for OSR calculation will be Therefore, assuming a value for OSR, for example OSR=10, and Cref=2.67Mbps for QPSK ½ the total capacity demand that can serviced is and the number of subscribers N can be estimated from (1).

Network dimensioning Network dimensioning is the process one follows with a view to estimating the hardware equipment needed for the network operation. This process is accomplished in three design steps: Coverage dimensioning is the first step in network dimensioning. It is used to determine the number of required BSs to cover the service area based on the cell radius R. Services dimensioning is the second step in network dimensioning. It takes into account the various services made available by the network operator and their corresponding data rates and estimates the required total subscribers’ capacity demand inside a cell of radius R. Capacity dimensioning is the final step in network dimensioning. The cell footprint must be sufficient to accommodate the requested capacity inside the BS cell. The number of sectors per cell required is determined.

Coverage dimensioning: design steps (1/2) The process is simple: the service area (km2) is divided by the cell footprint to produce the necessary points of presence (PoP) where a Base Station will be deployed. to calculate the cell footprint, a very significant step is to estimate the cell range. The cell range, R, will correspond to the distance from the BS that a minimum received SNR is achieved for the lowest MCS to be used by the network operator. Given the cell range R, the BS capacity (throughput) can be estimated, as we saw earlier, using MCS zones or OSR.

Coverage dimensioning: design steps (1/2) The cell range below is based on SNRQPSK = 4 (6 dB)

Coverage dimensioning: number of BSs Given the cell radius R

Services dimensioning Services dimensioning is the second step in network dimensioning. It takes into account the various services made available by the network operator and their corresponding data rates and estimates the required total subscribers’ capacity demand inside a cell. Determining the data density requirements for a particular region is a multi-step process as shown below:

Services dimensioning The capacity requirement is an important aspect of the network design and depends on the number of subscribers and type of data services offered. In some works, data density, expressed in Mbps per km2 is used to describe capacity requirements. In this case, the capacity demand per BS cell will be

Services dimensioning Most common categories of IP-network services are VoIP: estimate the required VoIP capacity per sector. This estimation is based on the required number of concurrent calls in a sector during the busy hour. Broadband data: broadband Internet is offered in terms of DL/UL peak information rate (PIR) in Mbps. The committed information rate (CIR) and the notion of Contention Ratio (CR) are used in network planning. Guaranteed bandwidth: is a form of data service which exhibits strict QoS requirements, and is usually provided to corporate customers for establishing VPN or as leased line equivalent.

Services Dimensioning: VoIP service Offered traffic (in erlangs) is related to the call arrival rate, λ, and the average call-holding time (the average time of a phone call), h, by: provided that h and λ are expressed using the same units of time (seconds and calls per second, or minutes and calls per minute). Busy Hour Traffic (in Erlangs) is the number of hours of call traffic during the busiest hour of operation of a telephone system. Blocking is the failure of calls due to an insufficient number of lines being available.  E.g. 0.03 mean 3 calls blocked per 100 calls attempted (which becomes the target probability of call blocking, Pb, when using the Erlang B formula).

Services Dimensioning: VoIP service Given that VoIP service operates analogous to PSTN, the active calls N can be estimated by utilizing the Erlang B Equation, where: Pb is the probability of blocking N is the number of identical parallel resources such as servers, telephone lines, etc. A = λh is the normalized offered traffic stated in erlang.

Services Dimensioning: VoIP service Example 1: Suppose a call center has N=10 phone lines, receives 480 calls per day, and the average duration of a call is 15 minutes. Since 15 minutes = 1/96 days, the number of Erlangs is (480)(1/96) = 5. (Computing Erlangs requires that call frequency and call duration be in the same units of time. You will get the same number no matter which units you use.) Thus, the probability that a call is blocked is GoS = (510/10!)/(∑10n=0 5n/n!) = 0.0183846. This means about 1.84% of the calls get dropped. As N becomes large, calculating GoS by hand becomes unwieldy, thus Erlang calculators must be used.

Services Dimensioning: VoIP service Example 2: Recursive iterations can be used to figure N from a given value of A and a desired value of GoS. Suppose that a call center receives 300 calls per hour, the average call duration is 5 minutes, and the center would like a GoS value of 0.025. First, we calculate the number of Erlangs as (300)(1/12) = 25, since 5 minutes is 1/12 of an hour. Then using an iterative procedure similar to guess-and-check, we find that the call center must have a minimum of 33 phone lines to achieve this Grade of Service. 

Services Dimensioning: VoIP service How to read the previous table: Erlang B table consists of Grade of Service (GoS) in X axis and No. of lines in Y axis. If our system capacity is 26.4 erlang and GoS is 2%, then we see the value of y axis where 26.4 or higher found below 2%. In the previous picture we can see we will be require 35 lines to satisfy condition. The selection of Erlang B equation is based on three important assumptions: Poisson call arrivals, fixed or exponentially distributed call holding times, and blocked calls are cleared.

Services Dimensioning: VoIP service The voice codec transforms the voice stream into data packets, however on top of the data packet a number of headers are attached prior to the air-interface. These headers are related to the Ethernet and Internet protocols. The required throughput is calculated by multiplying # of active calls N with the codec rate from Table 13.1.

Services Dimensioning: VoIP service Consider a simple example, where each customer has two telephone lines with 20 mE, 40mE, and 75mE traffic activity per line and Pb = 0.01. The overall Sector traffic activity A is computed by multiplying # of subscribers/sector x lines x traffic activity/line, Using Erlang B the number of active calls, N, is calculated. Example: A = 180 x 2 x 0.02 = 7.2 Erlangs Using Erlang B formula (see Erlang calculator below) we obtain N =14 lines.

Services Dimensioning: VoIP service The graph of N versus the number of subscribers in a sector is The required throughput (bandwidth) is calculated by multiplying # of active calls N with the codec rate (kbps) in Table 13.1.

Erlangs to VoIP calculators Erlang calculator http://www.erlang.com/calculator/erlb/ VoIP bandwidth calculator http://www.erlang.com/calculator/lipb/ This calculator can be used to estimate the bandwidth required to transport a given number of voice paths through an IP based network.   Reverse calculations are also possible.  This estimates the number of voice paths that can be transmitted though an IP network if the available bandwidth is known. Before a calculation can be performed, details of the voice compression scheme must be entered into the first two areas of the calculator.

VoIP bandwidth calculator For example, to work out how much bandwidth is required to transmit 10 voice paths (or channels) through an IP network using the G.723.1 (6.4kbps) coding scheme with 30ms packet duration, follow these steps: Use the Coding algorithm drop down list to select G.723.1 (MP-MLQ) 6.4 kbps compression. 30ms should automatically be selected as the Packet duration.  . Ensure that the Unknown radio button in the Bandwidth area is selected. Enter 10 into the edit box within the Voice paths area. Press the Calc. button.  After a short time, 171 should appear in the Bandwidth edit box, indicating that a bandwidth if 171kbps would be required.

Erlangs to VoIP Bandwidth Calculator The three variables involved in these calculations are Busy Hour Traffic (BHT), Blocking and Bandwidth. For example, to work out how much bandwidth is required to transmit 45 Erlangs of busy hour traffic through an IP network using the G.723.1 (6.4kbps) coding scheme with 30ms packet duration, follow these steps: Use the Coding algorithm drop down list to select G.723.1 (MP-MLQ) 6.4 kbps compression. 30ms should automatically be selected as the Packet duration.   Ensure that the Unknown radio button in the B/W (bandwidth) area is selected. Enter 45 into the edit box within the B.H.T. area. Press the Calc. button.  After a short time, 990 should appear in the Bandwidth edit box, indicating that a bandwidth if 990kbps would be required to carry the 58 voice paths required.

Services Dimensioning: Broadband Data Contention Ratio: This is defined as the ratio 1/CRi, where CRi equals to the maximum number of users that can be serviced using a dedicated channel for service i. For example, for Internet service we can put CR=25, that means that a channel of 768kbps can serve 25 subscribers, assuming that not all of them will need to have access of the full channel at the same time. This is a common practice employed by network designers. Different services have different QoS requirements that lead to different CR Residential class users can have CR 20-50 Business class users can have CR 4-10

Services Dimensioning: Broadband Data Dimensioning for Service Peak information rate, PIR, for each service class Committed information rate, CIR, for each service class Contention Ratio (CR) or over-booking rate (O) Assuming a number of different service classes, we have

Network dimensioning (1/2) Dimensioning for capacity and frequency planning After achieving coverage dimensioning, the system designer needs to carry out capacity dimensioning to ensure that the capacity offered to the customers satisfies the required capacity demand according to the number of subscribers, type of service used by them and data traffic in the network. Outcome of this process is the determination of the number of sectors per cell to address the need of required data rate for the subscribers using different services.

Network dimensioning (2/2) 1. The capacity of a sector will be the same as the capacity of the original cell using an omnidirectional antenna For this to happen each sector uses a different 3.5 MHz bandwidth, that is, different carrier frequency per sector. For example, a tri-sectored BS will need 3×3.5MHz =10.5 MHz Customer demand should be estimated based on average data rates and VoIP Committed Information Rate. Final step is to determine the number of required sectors, by dividing the number of customers per cell area with the number of customers per sector

Capacity of BS (BW=3.5 MHz) OFDM Bandwidth efficiency' CAPACITY ANALYSIS OFDM BW efficiency (b/s/Hz) 0,69 Modln+coding efficiency (b/s/Hz) 0,50 1,00 1,50 2,00 3,00 4,00 4,50 Overall PHY layer efficiency, (b/s/Hz) 0,35 1,04 1,38 2,07 2,76 3,11 User Data Rate, Mbps 1,21 2,42 3,63 4,84 7,26 9,68 10,89 (BW=3.5 MHz) OFDM Bandwidth efficiency' N_fft = Number of OFDM tones 256 N_data = Number of data tone 192 n = Sampling factor 8/7=1,152 Guard band efficiency (192*8/7) /256=0,864 Cyclic prefix guard time factor (Tg/Tb) 0,250 Guard time efficiency 1/(1+0,250)=0,8 OFDM Bandwidth efficiency factor (downlink) 0,864*0,8=0,691 OFDM Bandwidth efficiency factor (uplink) 0,691

Capacity of BS ID MCS SNR_min (dB) Throughput Mb/s 1 BPSK 1/2 6.4 1.21 QPSK 1/2 9.4 2.42 3 QPSK 3/4 11.2 3.63 4 16-QAM ½ 16.4 4.84 5 16-QAM 3/4 18.2 7.26 6 64-QAM 2/3 22.7 9.68 7 64-QAM 3/4 24.4 10.89

Capacity of BS if the cell radius R is selected to be 1.5 Km, then the above graph in the downlink for outdoor NLOS type A environment gives Using the above values the supported downlink throughput for outdoor NLOS category A environment with R=1.5Km is given by

Estimation of number of subscribers per BS For example, let us consider the case of a WiMAX operator that offers the following service classes (Agbinya, 2009): SME users, 5% of total number of subscribers, with PIR of 1Mbps with CR=10. SOHO users, 15% of total number of subscribers, with PIR of 512 Kbps with CR=10. residential users, 80% of total number of subscribers, with PIR of 1Mbps with CR=20. The average capacity per subscriber is defined as follows (Agbinya, 2009)

Estimation of number of subscribers per BS Assuming two service level agreements (SLAs) offered by the operator, for the first year, we can have the service mix shown below: which results in: SLA i Pi PIRi (Mbps) CRi Residential 80% 1.0 30 Business 20% 2.0 10

Estimation of number of subscribers per BS For single WiMAX BS with radio frequency bandwidth 3.5 MHz, operation frequency of 3.5GHz, and terrain category B: The average capacity supported by the BS with cell radius R=2.0 Km is given by 6.531 Mbps that is, 78 residential users and 19 business users can be supported by the BS capacity. MCS Mbps MCS zone (km) 64-QAM ¾ 10.89 0 – 1.0 64-QAM 2/3 9.68 1.0 – 1.1 16-QAM ¾ 7.68 1.1 – 1.4 16-QAM ½ 4.84 1.4 – 1.5 QPSK ¾ 3.63 1.5 – 2.0 QPSK ½ 2.42 2.0 – 2.2 BPSK ½ 1.21 2.0 – 2.6

Estimation of number of subscribers per BS Assuming that Qj is the percentage of the total population of the area that uses the j-th MCS level, demographic data is needed to estimate the Qj values depending on the area of installation. For example, a suburban area can be modeled as follows (these values are only indicative and geographical and demographic data are imperative to selecting the Qj values). MCS Qj 64-QAM ¾ 25% 64-QAM 2/3 20% 16-QAM ¾ 16-QAM ½ QPSK ¾ 15%

Estimation of number of subscribers per BS However, the above number of customers N can be multiplied by an OSR parameter, which usually takes a value between 1-2 (Agbinya, 2009), and accounts for the fact that some subscribers will use higher MCS levels. Assuming OSR=1.4 the BS can sign-up 109 residential users and 26 business users. It follows that the number of subscribers that can be serviced by the WiMAX BS for a given cell radius varies according to choice of values Qj (this will impact the system’s OSR) choice of values Pi (percentage distribution of residential and business users) supported data rates and CR per service class

References Agapio, P., et al. (2007) A model for WiMax coverage and capacity performance estimation. [Online] Available from web1.see.asso.fr/cdecrr07/papers/S5.3.pdf  Agbinya, J. I. (2009) Planning and optimization of 3G and 4G wireless networks. River Publications, Denmark. Smura, T. et al., (2008) Techno-Economic Analysis of Fixed WiMAX Networks.  In Zhang, Y. & Chen, H. H. (eds). Mobile WiMAX -- Toward Broadband Wireless Metropolitan Area Networks, pp. 363-385.