Online Spectrum Allocation for Cognitive Cellular Network Supporting Scalable Demands

State Key Laboratory of Virtual Reality Technology and Systems

Online Spectrum Allocation for Cognitive Cellular Network Supporting Scalable Demands

Jianfei Wang, Jinzhao Su, Wei Wu 2011-6-2


Outline• Motivation• Consideration• Algorithm Detail• Evaluation• Conclusion


Motivation• To Decrease Client Waiting Time– Online Spectrum Allocation

• To Take advantage of scalable property of request– Satisfaction Degree• Satisfaction degree is equal to ratio of allocated

bandwidth versus requested bandwidth


Consideration • Don’t reclaim client’s resource by force.– Reclaim will increase system’s network cost.– Reclaim will decrease client’s experience.

• High Spectrum Utility Rate– The higher is spectrum utility rate in busy time,

the better• Low denial of service

Spectrum Utility Rate

denial of service


System Model• Complete graph– Allocation among users of the same base station

• Coordinate Network– Users must get Base station’s permission


System Assumption– Arrival of users’ request• Possion process

– Users’ service time• Exponential distribution

– Queuing rules• FCFS

– Capacity of the system• Users’s amount must be less than N0


State Machine of System

r


Stabilization of Normal State• According to the assumption, the system in

the normal state can be modeled as queueing system, M/M/N0/N0.

• We can get the relationship among satisfaction degree , departure rate and arrival rate. 0* *

a

f wrw

0 0

0

0!0

k

kk

p k Np k

k N


Indentify State Transform• Boundary between Normal and Poor,

• Boundary between Rich and Normal,

• Target Frequency Ratio of Normal State,– From Frequency Utility Ratio’s aspect, the larger is f ,

the better.– From denial of service’s aspect, the lower is f, the

better.

1( ( ) | )n t nP N t k R

2( ( ) | )m t nP N t k R

poorf

richf

normalf


Estimation of Arrival Rate• From the pictrure, we can see that the arrival

rate has the seasonal propersty. So we use the seasonal arma model to estimate the arrival rate.

24

( ) ( )

( )(1 ) ( )( ) ( )

B z t

B B z tB a t


Estimation of Departure Rate• Maximum likelihood– The Customers who recently depart• The number of Customers who recently depart is• The service time of customer i is

– The customers who are active• The number of active customers is • The service time of active customers which has elapsed

is 1 1 * ( )1ˆ ( )

est

i activei N

est

t T N t

N N t

activeT

itestN

( )N t


Evaluation• Evaluation Data– Wifidog[7] which were collected from a large number

of free Wi-Fi hotspots in Canada for three years.• Evaluation Parameters– Target Frequency Utility Ratio of Normal State is 80%– Boundary between Normal and Poor and Boundary

between Rich and Normal are derived from training with part of wifidog data.


Evaluation ResultActive Consumers vs Satisfaction System Free Spectrum Statistics


Conclusion• we propose an algorithm for online spectrum allocation for

scalable demands in cognitive cellular network.• We introduce a concept of users’ satisfaction degree to make

good use of scalable property of demand.• To handle the issues produced by online property, we involve

queueing system and estimations of arrival and departure rate to balance the spectrum utilization and future demands.

• With theoretical analysis of system, we give the method of calculating parameters of system.

• At last we use collections of real wireless data to evaluate our algorithm.


Reference• [1]G.E.P.BoxandG.M.Jenkins. Timeseriesanalysis:forecastingand control.PrenticeHallPTRUpperSaddleRiver,NJ,USA,1994.• [2]M.BuddhikotandK.Ryan.Spectrummanagementincoordinated dynamicspectrumaccessbasedcellularnetworks.In

Proc.ofIEEE DySPAN,pages299–307,2005.• [3]L.Cao,L.Yang,X.Zhou,Z.Zhang,andH.Zheng.Optimus:SINR-DrivenSpectrumDistributionviaConstraintTransformation.In

New FrontiersinDynamicSpectrum,2010IEEESymposiumon,pages1–12. IEEE,2010.• [4]X.Cheng,C.Dale,andJ.Liu.Statisticsandsocialnetworkofyoutube videos.In

QualityofService,2008.IWQoS2008.16thInternational workshopon,pages229–238.IEEE,2008.• [5]L.Kleinrock.Queueingsystems,volume1:theory,1975.• [6]A.Lakhina,K.Papagiannaki,M.Crovella,C.Diot,E.D.Kolaczyk,and N.Taft.Structuralanalysisofnetworktrafficflows.

ACMSIGMETRICS Performanceevaluationreview,32(1):61–72,2004.• [7]MichaelLenczner,BenoitGrgoire,andFran?oisProulx.CRAWDAD tracesetilesansfil/wifidog/session(v.2007-08-

27).Downloadedfrom http://crawdad.cs.dartmouth.edu/ilesansfil/wifidog/session,August2007.• [8]J.Sachs,I.Maric,andA.Goldsmith.CognitiveCellularSystemswithin theTVSpectrum.In

NewFrontiersinDynamicSpectrum,2010IEEE Symposiumon,pages1–12.IEEE,2010.• [9]T.Schierl,T.Stockhammer,andT.Wiegand.Mobilevideotransmission usingscalablevideocoding.

IEEETransactionsonCircuitsandSystems forVideoTechnology,17(9):1204–1217,2007.• [10]J.Sjoberg,M.Westerlund,andA.Lakaniemi.Q.Xie,”RTPPayload FormatandFileStorageFormatfortheAdaptiveMulti-

Rate(AMR)and AdaptiveMulti-RateWideband(AMR-WB)AudioCodecs.Technical report,RFC4867,April2007.• [11]S.Sodagari,A.Attar,andS.G.Bilen.StrategiestoAchieveTruthful

SpectrumAuctionsforCognitiveRadioNetworksBasedonMechanism Design.In NewFrontiersinDynamicSpectrum,2010IEEESymposium on,pages1–6.IEEE,2010.


Any Questions?Thank you

Online Spectrum Allocation for Cognitive Cellular Network Supporting Scalable Demands

Documents

Transcript of Online Spectrum Allocation for Cognitive Cellular Network Supporting Scalable Demands