A NEW METHODOLOGY FOR ESTIMATING THE SPECTRUM

REQUIREMENTS WITH DATA TRAFFIC

Eun-Saem Yang, Hwa-Jong Kim

Division of Information Engineering and Telecommunications, Hallym University, Chuncheon, Korea

Department of Electronics and Computer Engineering, Kangwon National University, Chuncheon, Korea

Won-Seok Yang

IT Policy Research Group, ETRI, Daejeon, Korea

Keywords: Estimating the Spectrum Requirements, Bandwidth Estimation, Capacity Planning, Multimedia Data

Traffic, Mobile Network, WiBro.

Abstract: In this paper, we present a new methodology for estimating the spectrum requirements of next generation

mobile network with multimedia data traffic. In order to fully reflect the characteristics of mobile

multimedia data traffic in the spectrum requirement estimation, we take senven factors into account: self-

similarity of the data traffic, the layered structure of the data traffic, asymmetry of the data traffic between

uplink and downlink, engineered capacity considering QoS, regional simultaneous FA increase, uneven

traffic pattern among base stations, and handoff traffic overhead.

1 INTRODUCTION

Recently, along with the introduction of various

mobile access technologies, spectrum demand has

been rapidly increased. Therefore, an efficient

allocation of spectrum is becoming a principal issue.

In order to properly allocate spectrums to service

providers, we have to determine how much spectrum

is required for specific applications. Predicting the

spectrum requirement appropriately is essential for

the operators to guarantee the pre-defined quality of

service (QoS) and also helps the government to

license the proper number of service providers.

This study analyzes the existing studies on data

traffic, and presents core factors that must be taken

into consideration for spectrum requirements

estimation in mobile network with data traffic.

Well-known methodologies for spectrum

requirement estimation for mobile services include

the Rec. ITU-R M.1390 (1999) and the Report ITU-

R M.2023 (2000), which designed especially for

IMT-2000. These approaches are based on a circuit-

switching model, which mostly assumes voice traffic.

However, it is known that data traffic has a self-

similarity characteristic unlike voice traffic

(Crovella, M. E, 1997) (Leland, 1994). Therefore,

data traffic cannot be modeled accurately with

Poisson process or Busy Hour Session Attempts

(BHSA), which have been widely used for voice

traffic modeling.

Furthermore, it is known that the self-similarity

produces longer queue length and delay

(Grossglauser M. 1999) (Stallings, 1998). Hence,

this characteristic should be taken into account when

the service (system) capacity is calculated. This

paper adopted the “engineered capacity,” which

reflects the QoS in calculating the system capacity.

We assumed a “ regional simultaneous FA

increase” in order to avoid possible deterioration of

service quality due to hard handoff.

In the ITU-R model, the frequency efficiency of

a region is assumed to be constant in calculating the

spectrum requirement (ITU-R, 1999) (ITU-R, 2000).

However, the frequency efficiency of a region is not

constant, and this non-constant characteristic should

be considered in spectrum requirement estimation.

The paper is organized as follows. In Section 2,

we examine ITU-R's methodology and its limitation.

269

Yang E., Kim H. and Yang W. (2007).

A NEW METHODOLOGY FOR ESTIMATING THE SPECTRUM REQUIREMENTS WITH DATA TRAFFIC.

In Proceedings of the Second International Conference on Wireless Information Networks and Systems, pages 253-256

DOI: 10.5220/0002146702530256

 SciTePress

In Section 3, we describe the key factors in detail. In

Section 4, we propose a new methodology for

estimating spectrum requirements, and the

conclusion follows in Section 5.

2 ESTIMATING THE SPECTRUM

REQUIREMENTS IN ITU-R

There have been many studies for estimating the

spectrum requirement for mobile networks with data

traffic such as IMT-2000 and WiBro. Well-known

methodologies include the Rec. ITU-R M.1390

(1999) and the Report ITU-R M.2023 (2000), which

designed especially for IMT-2000. These

approaches are based on a circuit-switching model,

which mostly assumes voice traffic.

ITU-R M.1390 considers the BHSA for

estimating the Erlang traffic per user, assuming that

call arrival follows the Poisson Process, and using

exponential distribution model for session holding

duration. As the traffic for data services is bursty,

the activity factor must be taken into consideration.

For measurment of QoS, the Erlang-B formula

representing the call cut-off probability is used.

Moreover, the transmission rate of service channel is

set to average sector throughput, which is a physical

capacity. The spectral efficiency per cell assumes

equal for all cells. ITU-R M.1390 uses the method

described above to determine the final spectrum

requirements (ITU-R, 1999).

ITU-R M.2023 determines the spectrum

requirements for IMT-2000 based on the Erlang-C

model under the assumption that traffic follows the

Poison Process with consideration given to BHSA.

However, it is known that data traffic has a self-

similarity characteristic unlike voice traffic (Leland,

1994). Therefore, data traffic cannot be modeled

accurately with Poisson process or BHSA, which

have been widely used for voice traffic modeling.

Furthermore, it is known that the self-similarity

produces longer queue length and delay (Stallings,

1998). Hence, this characteristic should be taken into

account when the service (system) capacity is

calculated.

3 KEY FACTORS IN

ESTIMATING THE SPECTRUM

REQUIREMENTS

3.1 Characteristics of Data Traffic

3.1.1 Self-Similarity of Data Traffic

Many studies have reported that data traffic has self-

similarity (Crovella, M. E, 1997) (Leland, 1994).

The self-similarity of traffic implies that the present

traffic pattern has been influenced by past traffic

(Leland, 1994) (Willinger W, 1998). Because of the

self-similarity, data traffic cannot be modeled by a

Poisson process, which assumes that the current

state is independent of the past states. And the

BHSA also does not fit well for modeling data

traffic. ITU-R M.2023, however, did not consider

self-similarity because the spectrum requirement of

IMT-2000 was estimated by using the Erlang-C

model, which assumes Markovian call arrivals and

holding times.

In this paper, self-similarity of data traffic is

considered in spectrum requirement estimation

because future mobile service will deliver much data

traffic as well as voice traffic.

3.1.2 Layered Characteristic of Data Traffic

A typical traffic model used for the Internet has a

layered traffic structure of packet, packet call, and a

session layer appear (TR45.5, 1998). It also shows

the asymmetry of traffic flow between uplink and

downlink.

ITU-R M.2023 used the “activity factor” in order

to obtain Erlang traffic in modeling such layered

traffic. However, this scheme does not model data

traffic accurately because the Erlang model works

well for voice traffic. In order to alleviate this

discrepancy, a heuristic approach such as the

simulation approach was introduced (Yong-Joo

Chung, 2003).

3.2 Engineered Capacity Considering

QoS

In estimating the spectrum requirement, average

sector throughput has been widely used as the

physical layer capacity (ITU-R, 1999). However, the

average sector throughput usually does not take into

consideration QoS issues such as delay. For

example, delay variation over utilization for voice

and data traffic (with self-similarity) is shown in

Figure 1 (Stallings, 1998). In order to handle the

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

270

data traffic in estimating the spectrum requirements,

we need to consider the “engineered capacity,”

which takes delay into consideration

Delay

Utilization

Engineering limit

Under- estimate

Self- similar traffic Voice traffic

Delay

Utilization

Engineering limit

Under- estimate

Self- similar traffic Voice traffic

Figure 1: Delay of Self-similar Traffic over Utilization.

3.3 FA Increase Scheme

An example of WiBro spectrum allocation is shown

in Figure 2 (MIC, 2004). The unit of spectrum

allocation is called the frequency assignment (FA).

1FA 2FA 3FA 4FA 5FA 6FA 7FA 8FA 9FA

ISM

(WLAN)

A(30MHz)

B(30MHz) C(30MHz)

G/B

10MHz

2,400MHz2,3902,300MHz

3MHz 1.5MHz 3MHz1.5MHz

1FA 2FA 3FA 4FA 5FA 6FA 7FA 8FA 9FA

ISM

(WLAN)

A(30MHz)

B(30MHz) C(30MHz)

G/B

10MHz

2,400MHz2,3902,300MHz

3MHz 1.5MHz 3MHz1.5MHz

Figure 2: WiBro Spectrum Bands.

We assumed a “regional simultaneous FA

increase” in order to avoid the possibility of hard

handoff. The rationale for a “regional simultaneous

FA increase” is as follows. When the number of FAs

for each cell is not the same in a region, a hard

handoff may occur if a different FA is assigned to

the mobile terminal during handoff. A disconnection

during a hard handoff, however, would deteriorate

the QoS. We therefore assumed that all the FAs of

base stations are extended by the same number

(simultaneously) when the number of overloaded

base stations in a “handoff region” exceeds a

threshold.

3.4 Uneven Traffic Pattern among Base

Stations

Figure 3 shows a sample distribution of the offered

load of base stations in a region. The offered load

varies because of the differences in subscriber

mobility, subscriber density, and the users’

dispositions.

10%

15%

20%

25%

0% 20% 40% 60% 80% 100%

섹터

가동률

Offered Load at Base Station

Probability

10%

15%

20%

25%

0% 20% 40% 60% 80% 100%

섹터

가동률

Offered Load at Base Station

10%

15%

20%

25%

0% 20% 40% 60% 80% 100%

섹터

가동률

Offered Load at Base Station

Probability

Figure 3: Offered Load at Base Station.

The fact that the offered load varies for the base

station implies that spectral efficiency is not fixed

for each base station even if service properties are

equal. In this paper, the uneven traffic pattern among

base stations is considered to obtain the regional

operation ratio.

3.5 Handoff Traffic Overhead

Traffic overhead due to handoff in the CDMA

system is known to be about 30% of the total traffic.

It is noted that the TDMA or FDMA system may

also generate handoff overhead traffic in some cases.

In this paper, we considered the handoff overhead in

the estimating spectrum requirements.

4 PROPOSED MODEL

The new model for estimating the spectrum

requirements of a mobile network with multimedia

data traffic is proposed. The symbols used in the

model are summarized in Table 1.

Table 1: Symbols for the Proposed Model.

N Number of local subscribers

h Handoff traffic rate

U , D Uplink, Downlink

)(

⋅

T Traffic per subscriber (uplink/downlink)

)(

⋅

T Total traffic in a region (uplink/downlink)

)(

⋅

r Share of base stations(BS) (Ratio of physical

capacity over engineered capacity)

S Total number of sectors of BS in a region

)(

⋅

C Average sector throughput of BS (physical

capacity)

)(

⋅

C Engineered capacity of BS considering QoS

)(

⋅

C Total engineered capacity of BS in a region

(uplink/downlink)

m Base station capacity margin

)(

⋅

A Regional operating ratio (uplink/downlink)

f Standard value for capacity increase

The total uplink or downlink traffic in a region is

determined by the number of users, traffic per user,

and handoff rate:

A NEW METHODOLOGY FOR ESTIMATING THE SPECTRUM REQUIREMENTS WITH DATA TRAFFIC

271

DUihiNTiT

, ),1)(()( =

(1)

The total engineered capacity of the base stations

in a region is defined as the product of the

engineered capacity of a base station and the total

number of sectors.

DUimiSCiC

, ),1)(()( =

−

(2)

In Eq (2), m is used for the marginal capacity of

the base station, and for the engineered capacity of

base stations considering the QoS,

)(

⋅

C is

expressed as follows:

DUiiCiriC

, ),()()( ==

From Eqs (1) and (2), the regional operating ratio,

)(iA is defined as follows:

DUiiCiTiA , ,)()()( ==

(3)

The regional simultaneous FA increase method

can be expressed as:

fUA ≥)( or fDA ≥)( , increase the regional FA

If we assume that traffic is uniformly distributed

for each base station and the loads are the same,

becomes one. However, as shown in Figure 3, the

load is not the same for all base stations. Therefore,

the uneven pattern of traffic distribution in a region

makes

f less than one.

Following the above process, the required

number of FAs in a region can be calculated. It is

noted that the spectrum required for a specific

service should be chosen as the maximum value

among all regional spectrum requirements.

5 CONCLUSIONS

In this paper, we investigated key factors that may

affect the estimating of spectrum requirements of

next generation mobile network with multimedia

data traffic. First, the self-similarity and layered

structure of data traffic was considered.

Characteristics of data traffic were reflected via

simulations. And the asymmetry of data traffic

between uplink and downlink was considered to

include both types of traffic. Second, engineered

capacity based on the QoS (e.g., delay) was used as

the system capacity. Third, we assumed a “regional

simultaneous FA increase” in a region in order to

avoid the possibility of deterioration of service

quality. Fourth, uneven traffic patterns among base

stations in a region were considered, and finally

handoff overhead traffic was taken into

consideration.

The next study will be showed some numerical

examples of the proposed methodology applied to

communication technologies. In addition, we need to

study in more detail how each of the parameters will

change when the individual service environment

such as service type and characteristics changes.

ACKNOWLEDGEMENTS

This work was supported in part by MIC, Korea

under the ITRC program (C1090-0603-0035)

supervised by IITA.

REFERENCES

Crovella, M. E. and Azer Bestavros, 1997. Self-Similarity

in World Wide Web Traffic: Evidence and Possible

Causes, IEEE/ACM Transactions on Networking, Vol.

5, No. 6, pp. 835-846.

Grossglauser M., and Jean-Chrysostome Bolot, 1999. On

the Relevance of Long-Range Dependence in Network

Traffic, IEEE/ACM Transactions on Networking, Vol.

7. No. 5, pp. 629-640.

ITU-R, 1999. Methodology for the calculation of IMT-

2000 terrestrial spectrum requirements, ITU-R M.1390.

ITU-R, 2000. Spectrum requirements for international

mobile telecommunications-2000 (IMT-2000), ITU-R

M.2023.

Leland, W. E., M. Taqqu, W. Willinger, and D. V. Wilson,

1994. On the self-similar nature of Ethernet traffic

(extended version), IEEE/ACM Transactions on

Networking, Vol. 2, No. 1, pp. 1-13.

MIC(Ministry of Information and Communication), 2004.,

Spectrum bands of WiBro (written in Korean),

no.2004-53.

Stallings, W., 1998. High-Speed Networks; TCP/IP and

ATM Design Principles, Prentice Hall.

TR45.5, 1998. cdma2000 RTT Candidate Submission to

ITU-R, U.S. TG8/1.

Willinger W., Vern Paxson and Murad S. Taqqu, 1998.

Self-Similarity and Heavy Tails: Structural Modeling

of Network Traffic, Working Paper.

Yong-Joo Chung, Chun-Hyun Paik, Hu-Gon Kim, Taek-

Jin Choi, Won-Suk Yang, Heung-Suk Hwang, 2003.

Traffic Modeling and Performance Analysis of Mobile

Multimedia Data Services, Journal of Korea

Operations Research and Management Science

Society, No.28, vol.2, pp.139-155.

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

272