Dynamic Spectrum Access in Cognitive Radio Using Markovian Queuing Model

(1)

Volume 4, No. 3, March 2013 (Special Issue)

International Journal of Advanced Research in Computer Science

RESEARCH PAPER

Available Online at www.ijarcs.info

On 19-20 Feb 2013 Organized by

2nd SIG-WNs, Div IV & Udaipur Chapter , CSI , IEEE Computer Society Chapter India Council , IEEE Student Chapter Geetanjali Institute of Technical Studies, Udaipur, Rajasthan, India

Dynamic Spectrum Access in Cognitive Radio Using Markovian Queuing Model

Pinki Yadav

Department of Electronics and Communication Engineering Faculty of Engineering and Technology,

Mody Institute of Technology & Science (Deemed University) Lakshmangarh, Dist. Sikar, Rajasthan,

Pin-332311, India. [email protected]

Partha Pratim Bhattacharya

Department of Electronics and Communication Engineering Faculty of Engineering and Technology,

Mody Institute of Technology & Science (Deemed University) Lakshmangarh, Dist. Sikar, Rajasthan,

Pin-332311, India. [email protected]

Abstract: Today there is huge demand for radio spectrum in wireless communication networks. Fixed spectrum policy is a major problem in solving spectrum crisis. Dynamic spectrum access is one method which gives the solution by allocating the available limited radio spectrum band between different licensed and unlicensed users. Dynamically accessing the unused spectrum is known as dynamic spectrum access (DSA) which becomes a promising approach to increase the efficiency of spectrum usage. This allows unlicensed wireless users (secondary users) to dynamically access the licensed bands from legacy spectrum holders (primary users) on an opportunistic basis. In this paper, DSA approach by using hierarchical access model is proposed with Markovian Queuing model for centralized architecture. The blocking probability and mean access delay for infinite and finite users are also found out. Results are then compared.

Keywords: Dynamic spectrum access; Primary users; Secondary user; Markovian Queuing model; blocking probability

I. INTRODUCTIONTODYNAMICSPECTRUM ACCESS

National regulatory bodies provide the radio spectrum band for utilization in different wireless services. In the U.S., the main authority for radio spectrum regulation is the Federal Communication Commission (FCC). The FCC’s spectrum policy gives the actual spectrum usage measurements. Radio spectrum is not fully utilized often and so the proper way for utilization of radio spectrum is to use dynamic spectrum access technique which is adopted in cognitive radio.

Dynamic spectrum access is a technique by which a radio system dynamically adapts to available spectrum holes with limited spectrum use rights, in response changing circumstance and objectives: interference created changes the radio’s state, changes in environmental constraints [1]. The main objective of the DSA is to overcome two types of interference concern: harmful interference caused by malfunctioning device and harmful interference caused by malicious users. Dynamic spectrum management is also referred to as dynamic spectrum access. DSA that was first demonstrated in 2006 by the Defense Advanced Research Project Agency (DARPA) and Shared Spectrum Company (SSC) of Vienna, VA [2]. Dynamic spectrum access overcomes the limitation of fixed frequency spectrum assignment and management. Dynamic spectrum access models can be broadly categorized under three models as dynamic exclusive-use, open sharing model, and hierarchical access model as shown in Fig. 1 [3].

Figure 1. Dynamic Spectrum Access

A. Dynamic exclusive use model:

Dynamic exclusive use model maintains the basic structure of the current spectrum regulation policy. Spectrum bands are licensed to services for exclusive use. The main objective of the dynamic exclusive use model improves the flexibility and spectrum efficiency. It manages spectrum in finer scales of time, space and frequency and use dimensions. There are two types of dynamic exclusive use model [3]:

a. Spectrum property rights: In spectrum property

rights licenses are allowed to sell and trade spectrum and to freely choose the technology. The economy and market will thus play a more important role in driving toward the most profitable use of this limited resource. Licenses have the rights to lease or share the spectrum for profit, such sharing is not mandated by the regulation policy.

b. Dynamic spectrum allocation: Dynamic spectrum

(2)

B. Open sharin

Open sharing m model. This m users as the ba model is supp wireless service scientific and m

ypes [4]: Unco Private-commo

a. Uncontro

open spec c. Private-co

technolog can be simu licensed) user users can oppo t is not occup There are also and Spectrum u

a. Spectrum

actively e

b. Spectrum

approach

davet al, Internat

All Rights Reserve nd temporal tra

Spectrum is use in a giv s allocation va urrent policy.

ng mode:

model is also odel employs asis for manag ported from th

es operating in medical (ISM ontrolled- comm ons.

lled-commons

ctrum access. no entity has

band. It is m t.

d-commons: Th

e tragedy of orm of order

ommons: Sp

gy and protoco may receive a ransmission p access the spe

cal access mod

ccess model [5 ultaneously s r and seconda ortunistically ac pied or fully u two types of underlay.

overlay: The

xplored in a go cated by Mito stic exploitat

“gaps” in spa e Radio will

band, which a certain time a trum bands d is model is s ows primary an y users can us y communicati

o relay prim premises for mple of spec nd OFDMA sy

underlay:

imposes const secondary use floor of prim ed signals ove

econdary user ge high data

ion power sho se, the assum users transmit

tional Journal of

d _CONFEREN

affic statistics allocated to en region and aries in a much

called spectru open sharing ing a spectral he phenomenal n the unlicense ) radio band. mons, Manage

s: This is also

When a spect exclusive lice maximum tran

his represents commons by or structure

ectrum owne ol for the CR a command fro parameter). CR

ccess the radio utilized by pr f model; Spect

e spectrum ov oing DARAP X ola targets fo tion of whi atial-temporal have to ident are not used and location an dynamically to shown in Fig.

nd secondary t se part of thei ion and remain mary transmis an overlay sy trum overlay ystem [7].

The spectrum traints on the ers so they op mary users. B er a wide freq s can potential a rate with own in Fig. 2 mption is ma all the time; th

Advanced Resea

NCE PAPER of different services for d at a given h faster scale

um commons among

peer-region. This l success of ed industrial,

It has three ed-commons,

o referred as trum band is ensed to the nsmit power

an effort to imposing a of spectrum

er specifies user access. om spectrum R user may

dio spectrum en primary d) user. CR o spectrum if

imary users. trum overlay

verlay model XG program r aggressive te-space or

domain [4]. tify the idle

by licensed nd use those o unlicensed

2 (a). This transmission.

ir power for n part of the ssion, these ystem model

is TDMA,

m underlay transmission perate below By spreading quency band lly achieve a very low 2 (b). In the ade that the

his approach

arch in Computer

pectrum white

Figure 2. Spectru

ctrum overlay on by SU in restriction on th h cases, the lic

limiting interf um overlay a d to as opportu een that availa queue model ion in cognitiv

II. Q

gle station queu of finite or infi

as shown in m is also referre A server can se

in a “busy or the arrival of mer is buffered ble, and wait

tly service dep cted for service

fferent types o e-time, queuin s can be class d by a standar ndall and it is ll’s Notation Kendall’s nota

ntary Queuing ine and is sow ymbols, i.e. A ost important p

igure 3. A Stand

pecial Issue, Marc

l C f

on detection space [3].

um overlay and un

approach relie the network a heir transmissi censed spectrum

ference to PU approach is u unistic spectrum ability is increa

is chosen for ve radio.

QUEUINGSY

uing system c finite size and o n Fig. 3. An d to as a servic erve only one c “idle” state. If

the customer, d, assuming t for its turn. parts, one of th

e according to of arrival patter ng discipline

ified by a que rd notation. It

named as Ke

ation is widel g system A/B wn in Table I. M A/B/X are ment

parameters to

dard Graphical Not

ch 2013, 120-125

n and utilizati

nderlay approaches

es on spectrum and does not i ion power. The m is open to t U. In this pap used which i m access (OSA

ased with queu Dynamic Spe

YSTEM

onsists of a qu one or more id

elementary qu ce station or sim customer at a t f all servers ar , the newly ar that buffer sp

When the cu he waiting cust

a queuing dis rn, distribution and the numb euing system

was proposed ndall’s Notatio

ly used to de B/X/Y/Z in qu Most of the tim tioned since th represent a qu

tation for the queu

(3)

Parameters

The rule that a from the queue of the current Commonly use

c.

Priority-service, Se

d. RSS - Ran e. FQ - Fair

priority, b way.

III. P

This model i Markovian que

nteresting feat both infrastru centralized net allocation to in master or clust base station

nfrastructure b scheme is an queuing netw distributed que other as a se equivalent mod s as shown in F queue where M denotes single BAQ) is mode Here G is gene of requests fro system. S deno available with

davet al, Internat

All Rights Reserve l situations cus am (exponenti times are ind ervice time distributed. Th servers to proc tomers. Practic ueuing system

TableI. Kenda

Refers To

nter-arrival distrib of customer

Service Time Distribution

Number of Paral servers

Maximum numbe ustomers allowed

queue

Scheduling discipli queuing strateg

a server uses to e when the ser

customer is ed queue discip Service reque st-serve basis rrives when th ocks the reques Service request

e manner. According to ervice requests ndom Selection r queuing does but serve the

PROPOSEDQ

is unique in euing for Cog ture of our mo ucture based

tworks. A re ntended users ter head and b

(BS) in based network nalyzed using ork model w ues is modeled emi Markovia del of a networ Fig. 4. In this m M means Mark

server and ba eled as an M/G eral distribution om both PU otes the availab h head and N

tional Journal of

d _CONFEREN

stomers arrive al inter-arrival dependent and can be deter ere may be a s cess the service cally the numb

is limited.

all’s Notation

Ex

o select the ne rver is finished called queuing plines are:

sts are served . In FCFS, i he queue is f st.

s are served in

o the importa s are served. n for Services. s not consider service reque

QUEUINGMO

sense based nitive radio n odel is its ada and infras ecord of spec is to be perfo by the fixed c

infrastructur ks respectively g distributed where one o d as M/M/1 qu an M/G/S/N rk of queues fo model SUQ is kovian distribut andwidth alloc G/S/N loss que

n as there may and SU ente ble of unoccup N denotes lim

Advanced Resea

NCE PAPER according to l times) and d identically

rministic or single server e requests of

ext customer d the service g discipline.

d on a

first-d first-distributefirst-d networks. An

aptability for structure-less ctra and its

rmed by the controller or re-less and y. The MAC cation queue uing system. y be a variety ering to this pied channels

mited BAQ

arch in Computer

ty. A special ca r of available AQ becomes a s is Poisson an ential for a q vian queuing of stochastic

process of s ts to be proc r the requests e process [10].

Figure 4. Queui

ueue stacking dary user que ng this queue ar ) basis. At any ed to the SU ts from SU an el. Thus wh

ncies to PU an are summed to

The queue so ion queue (BA

rformance Ana

rlang B formul (1878-1929) a [11]. It is us ng probability

g model and i secondary use sumption that



is less than es stable and mean rate , du

updated with idth requests nsed users, ind

rates and idth allocation

rate at BAQ i ty for BAQ is ean service dur t by CR that fi ed, blocking p Erlang-B form

PB

pecial Issue, Marc

l C f

ase when queu channels in th an M/M/S/S q nd the service t queuing mode model. These processes, c ervice request essed, queuing are selected

ing model for DSA

all entries of eue (SUQ) an formed is refe AQ) [10].

alysis using Er

la was develop and is also kn sed for infinite . Erlang mode is based on qu r queue) is of the traffic int an unity (i.e.

the output pro ue to Bruke’s t

CR engine ca of both lic dependent Poi

are summ n queue BAQ

is + .

= + ration for BAQ finds all the ch probability PB

mula as given b

ch 2013, 120-125

ue size is same he system, i.e queue. If the g discipline i to be served

A in CR Network

SUs is referre nd all the re first come firs bandwidth nee considers bo ho needs it li ing a numb ival rates of bo frequencies wi rred to as band

rlang-B model

ped by Anger K nown as Erlan e users and fin el is also kno ueuing theory. f M/M/1 type. tensity of this

<1Erlang), ocess is still P theorem. The apabilities to censed as w isson processe hannels with H

is given by th below in (1)[10

(4)

Pinki Yadavet al, International Journal of Advanced Research in Computer Science, 4 (3) Special Issue, March 2013, 120-125

2nd SIG-WNs, Div IV & Udaipur Chapter , CSI , IEEE Computer Society Chapter India Council , IEEE Student Chapter Geetanjali Institute of Technical Studies, Udaipur, Rajasthan, India Let N1 and N2 denote the mean number of messages in

queues SUQ and BAQ respectively. According to the M/M/1 classical theory, we have for queue SUQ [10]:

N1= (2)

According to the M/M/S/S classical theory, we have for queue BAQ:

N2= 1 (3)

Let T1 and T2 denote the mean delay experienced by a

message in crossing SUQ and BAQ, respectively in (4) from the little theorem:

T1 = and T2= (4)

To determine the total mean access delay T, we have considered two different cases depending on the input of the message in our system as follows:

Case a. SU messages arriving at SUQ from outside the

system. The probability that these messages arrive at SUQ from outside the system is given as

⁄  with a mean message delay TD1 given as (5):

TD1=T1+ 1 T2 (5)

Case b. PU messages arriving at BAQ from outside the

system. This situation occurs with probability ⁄  with the mean message delay TD2 given as

(6):

TD2=PB 0+ (1-PB) T2= (1-PB) T2 (6)

Thus, T can be obtained as in (7):

T= [T1+ (1-PB) T2] + [(1-PB) T2]

T= T1+ (1-PB) T2 (7)

By means of expressions of T1 and T2 using (4) and N1

and N2 using (2) and (3), we obtained the overall access

latency, T as follows in (8):

T=

T= (8) The equation defines the overall delay encountered by any SU to access a channel for transmission of its own voice/data. Thus, T represents the time interval between the request initiated by the SU and the request granted the required bandwidth.

TableII. Parameter Used in Calculation

Notation Parameter

Poisson arrivals with a rate λ

Exponentially distributed service time with a rate

Traffic intensity

Blocking probability PB

Total mean delay T

Blocking probability from (1) is plotted in Fig. 5. Fig. 5 shows the relation between the numbers of channels (s) and blocking probability (PB) for different values of

traffic intensity (SU). So it is clear from the figure that

when number of channel increases, blocking probability decreases and when the traffic intensity (SU) increases, blocking probability also increases.

Figure 5. Blocking probability (PB) as a function of different number

of channel(s)

Figure 6. Mean access delay (T) against traffic intensity of BAQ

From (8), total access delay against blocking probabilities of 5%, 10%, 30%, 50%, 75% and 95% are plotted in Fig. 6 for variation in traffic intensity of BAQ as 0.2, 0.4, 0.6, 0.8. The plot shows that for any given traffic intensity of BAQ, T increases with increase in traffic intensity of BAQ while decreases for an increase in the blocking probability.

B. Performance Analysis using Engset formula:

The Engset formula was developed by Tore Olaus Engset. Engset is used with “Finite sources” when blocked call is cleared [11]. The Enset loss formula applies to situations where the number of customers is small relative to the number of available channel [12]. SU type M/M/1/L queue used, where L is the size of waiting room. The case of the M/M/1/L queue represents the queue in which the arrival and departure rates are not constant. The customer arrival rate is λ1 is

(5)

2nd SIG-WNs, Div IV & Udaipur Chapter , CSI , IEEE Computer Society Chapter India Council , IEEE Student Chapter Geetanjali Institute of Technical Studies, Udaipur, Rajasthan, India sources. The number of customers is assumed to be M,

the number of channel S and the blocking probability PB

[12].

The Engset loss formula gives the blocking probability for the case M>S as follows

PB=

∑ (9)

Figure 7. Blocking probability(PB) as a function of different number

of channel(s)

Blocking probability from (9) is plotted in Fig. 7. It shows the relation between the number of channels (s) and blocking probability (PB) for different values of

traffic intensity (SU). It is clear that when number of channel increases, blocking probability decreases. It is also analyzed that when the traffic intensity (SU) increases, blocking probability also increases. From equation (8), total access delay against blocking probabilities of 5%, 10%, 30%, 50%, 75% and 95% are calculated for variation in traffic intensity of BAQ as 0.2, 0.4, 0.6, 0.8 and plotted in Fig. 8. The plot shows that for any given traffic intensity of BAQ, T increases with increase in traffic intensity of BAQ while decreases for an increase in the blocking probability.

Figure 8. Mean access delay (T) against traffic intensity of BAQ

C. Performance comparison between Erlang-B and Engset formula:

Comparison between the results obtained using Erlang-B and Engset formula are shown in Fig.9. For equal traffic intensities, the engset formula gives a lower blocking probability than the Erlang formula and tends to the later when the user population grows to infinity. Regarding the computational complexity the engset model is much more simple than the bidirectional markov chain used in Erlang formula.

Figure 9. Comparison between Erlang-B and Engset model

IV. CONCLUSION

Dynamic Spectrum access is a promising approach to utilize the unused spectrum by the primary user. In this paper, a centralized architecture coexisting with licensed users and unlicensed users is used for Markovian queuing model. Blocking probability and mean access delay for infinite users are calculated using Erlang-B formula and for finite users using Engset formula. Results obtained are found to be satisfactory. The results are then compared. For equal traffic intensities, the Engset formula gives a lower blocking probability than the Erlang formula and tends to be equal when the user population grows to infinity. The future work will include separate queue for high and low priority users.

V. ACKNOWLEDGMENT

The authors want to acknowledge the facilities provided by Mody Institute of Technology & Science (Deemed University), Lakshmangarh, Rajasthan for supporting the present work.

VI. REFERENCES

[1] Przemyslaw s Pawelczak, Gerard J.M Janssen and R.Venkatesha Prasad, “Performance Measures of Dynamic Spectrum Access Networks” IEEE Globecom, 28 November 2006 pp.1-22..

[2] http://www.mwrf.com/.

[3] Qings Zhao and Brian M. Sadler, “A Survey of Dynamic Spectrum Access” IEEE Signal Processing Magazine, May 2007 pp.79-89.

0 0.2 0.4 0.6 0.8 1

2 3 4 5 6 7 8 9 10 11 12 13 14

Blocking

Pr

oba

b

ili

ty

Number of channels

(6)

2nd SIG-WNs, Div IV & Udaipur Chapter , CSI , IEEE Computer Society Chapter India Council , IEEE Student Chapter Geetanjali Institute of Technical Studies, Udaipur, Rajasthan, India

[4] Anita Garhwal, Partha Pratim Bhattacharya,

“Dynamic Spectrum Access in Cognitive Radio: a brief review”, International Journal of Computer Application in Engineering Sciences, Special Issue on Computer Networks & Security, Vol.1 December 2011, pp. 149-153.

[5] Xavier Gelabert, “Cognitive Radio Enabling

Opportunistic Spectrum Access (OSA): Challenges and Modeling Approaches” UPF-DTIC Weekly Research seminar 28 October 2010.

[6] Sudhir Srinivasa and Syed Ali Jafar, “The

Throughput Potential of Cognitive Radio: A Theoretical Perspective” IEEE Communications Magazine, May 2007 pp.73-79.

[7] Danda B. Rawat and Gongjun Yan, “Spectrum

Sensing Methods and Dynamic Spectrum Sharing in Cognitive Radio Networks: A Survey” International Journal of Research and Reviews in Wireless Sensor Networks Vol. 1, No. 1, March 2011.

[8] www.eventhelix.com.

[9] Indika Anuradha Mendis Balapuwaduge,

“Performance Evaluation of Channel Aggregation Strategies in Cognitive Radio Networks with Queues” May 2012.

[10] Prabhjot Kaur, Arun Khosla and Moin Uddin,

“Markovian Queuing Model for Dynamic Spectrum Allocation in Centralized Architecture for Cognitive Radios” IACSIT International Journal of Engineering and Technology, Vol.3, No.1, February 2011 ISSN: 1793-8236.

[11] Richard parkison, “Traffic engineering Techniques in Telecommunications” copyright Infotel Systems corporation pp.1-29.

[12] Moshe Zukerman, “Introduction to Queuing