Enhanced DHCP for the Fast Retrieval of theSpectrum Map for White Space Applications

Seungil Yoon1, Kyutae Lim1, and Jongman Kim2

1 Georgia Electronic Design Center, School of Electrical and Computer Engineering,Georgia Institute of Technology, Atlanta, Georgia 30332, Email: {syoon7, ktlim}@ece.gatech.edu

2 KORUS Research Center for Informersive Systems, School of Electrical and Computer Engineering,Georgia Institute of Technology, Atlanta, Georgia 30332, Email: jkim@ece.gatech.edu

Abstract—As white space, unused local TV spectrum, willbe allowed for the broadband Internet access, we can usewhite space network as the supplementary network of licensedand/or unlicensed access networks such as cellular and WiFiTM

networks. Thus, after setting up the connection over a cellularor WiFiTM system, mobile devices equipped with a white spaceinterface will connect to a radio resource management server toretrieve a location-based list of open white space channels, “aso-called spectrum map”. This paper aims to reduce a signalingoverhead for the setup of a cellular or WiFiTM connection andthe retrieval of the spectrum map. To achieve the goal, we suggestembedding the retrieval operations of the spectrum map insidethe Dynamic Host Configuration Protocol (DHCP) operations forthe IP assignment during the connection setup of the cellular orWiFiTM networks. The theoretical analysis demonstrates that ourproposal, the fast provision of the spectrum map, can contributeto accommodating one or more white space devices when thetotal number of open white space channels is 30.

Index Terms—Dynamic Host Configuration Protocol, DHCP,Location-based list, White Space, Handover

I. INTRODUCTION

White space devices (WSDs) operating on white space,unused local TV spectrum, for the broadband Internet accessconduct spectrum sensing that aims to find spectrum holes,unused white space channels. They continue to perform spec-trum sensing while they successfully operate on their operatingchannel chosen among unused channels according to spectrumsensing results. With this continuity of spectrum sensing, theyare capable of immediately switching their operating channelto the new one, which is marked as unoccupied according tothe latest spectrum sensing results, when any primary userssuch as TV broadcasting services suddenly appear on theiroperating channel. However, because of the imperfect accuracyof spectrum sensing and dynamic spectrum availability, theycan have a chance to select a wrong channel that is marked asunoccupied according to spectrum sensing results but occupiedby other WSDs or primary users in reality. To reduce apossibility of wrong detection in spectrum sensing, WSDsretrieve a location-based list of open white space channels, “aso-called spectrum map”, from a common database server orradio resource management (RRM) server. With the list, theycan avoid spectrum sensing on the channels already occupiedor not allowed for the broadband Internet access.

For the retrieval of spectrum map, they need the Internetconnection, which means they have to have the previously

DHCP Server 1

RRM Server

DHCP Client

2. DHCP Discover

1. Connected to 802.11 WiFi or 3G

3. DHCP Offer

8. New connection to White Space Network

6. Query of retrieving the spectrum map

7. Reply with the spectrum map

RRM: Radio Resource Management

4. DHCP Request

5. DHCP Ack

9. Second DHCP operations for IP assignment of White Space Network

First DHCP opreations

Fig. 1. The general call flow of the handover from WiFi TM to White SpaceNetwork

connected wireless networks like cellular or WiFiTM networks.Thus, after they set up the Internet connection over cellularor WiFiTM networks, they connect to the spectrum map orRRM server for the retrieval of the spectrum map of a cur-rent location. Using either Global Positioning System (GPS)or network-based location mechanisms, they can determinetheir current location. With the setup of cellular or WiFiTM

networks and the retrieval of the spectrum map as depicted inFig. 1 in sequence, they have a possibility of having the secondtime of the Dynamic Host Configuration Protocol (DHCP)operations for the IP assignment as depicted in step 9 of Fig.1. Our main idea is to shorten the interval time between twoDHCP operations by embedding the retrieval of spectrum mapduring the first DHCP operations.

For the fast retrieval of spectrum map, we propose extendingthe DHCP protocol that is able to deliver to the DHCP clients,WSDs, the spectrum map during the setup of cellular orWiFiTM networks. During the DHCP operations for the IPassignment of cellular or WiFiTM networks, the DHCP serverconnects to the RRM servers to retrieve the spectrum mapand includes the retrieved spectrum map inside the DHCPmessages that will be delivered to the DHCP clients. In ourproposal, we also extend the spectrum map having not only

IEEE WCNC 2011 - Network

978-1-61284-254-7/11/$26.00 ©2011 IEEE 808

the list of the open white space channels but also the list ofavailable access networks. Similarly to the approach of [1], theextended spectrum map can assist the DHCP clients to selectthe least traffic load of an access network among multiplenetworks. The enhanced DHCP server can provide its locationinformation if the DHCP clients can not determine or offertheir location as [2] suggests. In addition, as [3] proposes, theenhanced DHCP server can dynamically manipulate the listin accordance with preferable access networks of the DHCPclients. However, in this paper, we present the modified setupprocedures for the fast retrieval of the spectrum map of openwhite space channels and define the new options that can beincluded in DHCP messages and used for the fast retrieval ofthe spectrum map.

This paper consists of the following sections: Section IIdescribes the conventional approach of extending the role ofthe DHCP server and presents our approach to shorten the totalsetup time for the retrieval of the spectrum map. Section IIIdescribes the analytical model for the performance evaluationof our proposal, and Section IV presents the analysis results.Section V describes the conclusions of this paper.

II. EXTENSION OF THE DHCP

At [2], the DHCP server only provides its location infor-mation not the spectrum map to the DHCP clients althoughthis provision can be useful to the clients when they can notdetermine their location under certain circumstances such asinside buildings. In other study [4] [5] [6], the DHCP serversare still limited to assist the session continuity, mobilitymanagement of mobile devices, or the awareness of securednetwork location. Our approach is to allow the DHCP server toassist the retrieval of the spectrum map directly from the RRMservers. Thus, compared to previous research, in our approach,the DHCP server acts as the main controller for assisting theDHCP clients to retrieve the spectrum map in advance. Wehave two approaches on the implementation of our proposal:Aggressive and cautious approaches.

A. Aggressive Approach

According to the message sequence of the typical DHCPoperation, since a DHCP Discover is usually the first message,the clients can include a new DHCP field option (ANMReqInd)that indicates their desire to retrieve a list of connectable accessnetworks (ANM:Access Network Map) in a current location.For this operation, we define an additional DHCP field for lo-cation information (LocInfo) in the DHCP Discover message,and the DHCP clients include their location information inthe LocInfo option. However, when the DHCP clients cannotoffer their location information, the DHCP clients include theonly ANMReqInd in the DHCP Discover, and in this case, theDHCP server uses their location information as the locationinformation of the DHCP clients. Fig. 2 depicts the new callflow of the retrieval of the spectrum map with the aggressiveapproach. The DHCP client that connects to any popularaccess network such as WiFiTM networks includes the ANM-ReqInd option and the LocInfo option in the DHCP Discover

Option 2.

Option 1.

DHCP Server 1 RRM ServerDHCP Client

3. Access Network Map (ANM) Retrieval

2. DHCP Discover

1. Connected to 802.11 WiFi or 3G

4. DHCP Offerwith ANM

5. New connection to White Space Network

6. DHCP Discover

7. DHCP Offer

8. DHCP Request

9. DHCP Ack

6. DHCP Request

7. DHCP Ack

RRM: Radio Resource Management

In option 1, the IP allocation is renewed since the client connects to the new sub-network.

In option 2, the client uses the offered IP in the DHCP Offer message on step 4 because the previous and new access networks are in the same sub -network

DHCP Server 2

Fig. 2. The new call flow of the DHCP in the aggressive approach.

message as depicted in step 1 of Fig. 2, and the DHCP server 1retrieves a list of connectable access networks, ANM, from theRRM server using the LocInfo option retrieved from the DHCPDiscover message or its own location information. The DHCPserver 1 sends the DHCP Offer message with the retrievedANM to the DHCP client, and then, the client performs thehandover to the new access network, white space network(WSN), as depicted in step 5 of Fig. 2. Afterward, the clienthas two options: restarting the IP assignment with the newDHCP server, the DHCP server 2 as depicted in steps 6-9 ofOption 1, or continuing the remaining procedures with the IPaddress offered by the DHCP server 1 as depicted in steps6-7 of Option 2. However, the DHCP client can have moreoptions like it still remains in the current access network oruses the retrieved ANM later for the mobility management asit moves to neighboring cells.

B. Cautious Approach

Instead of the DHCP Discover message, as depicted in step4 of Fig. 3, the DHCP clients can deliver the ANMReqIndin a DHCP Request message. The DHCP server will notsend the DHCP Offer, which is the response of the DHCPDiscover message, if the DHCP clients are not authorizedusers or devices. Thus, in this passive approach, they onlyrequest the ANM when they retrieve the DHCP Offer messagethat contains the assigned IP address so that they can retrievethe required ANM in a DHCP Ack message. Compared to aconventional procedure of the IP assignment on DHCP, theycan wait their settings of the IP address and configurationparameters until they retrieve the DHCP Ack message. In thiscase, they can switch their working access network to thenew one among available access networks in accordance withthe retrieved ANM. Since they use DHCP Request and Ackmessages generally for renewing or confirming the current IPaddress, they can use this defensive approach for the handover

809

Option 2.

Option 1.

DHCP Server 1 RRM ServerDHCP Client

5. Access Network Map (ANM) Retrieval

2. DHCP Discover

1. Connected to 802.11 WiFi or 3G

3. DHCP Offer

7. New connection to White Space Network

4. DHCP Request with ANMReqInd and LocInfo

6. DHCP Ack with ANM

RRM: Radio Resource Management

8. DHCP Request

9. DHCP Ack

DHCP Server 2

These two steps are intended to authorize the DHCP client before proving the ANM.

8.Normal DHCP operation from DHCP Discovery to Ack .

9. DHCP Release

Fig. 3. The new call flow of the DHCP in the cautious approach.

after they use a cellular connection for a while rather than animmediate handover from a newly connected WSN. As theaggressive approach does, they have two options. However, inthe first option, they send the DHCP Release message either inparallel during the new IP assignment with the DHCP server1 or shortly afterward when the new IP assignment is oversuccessfully as depicted in step 9 of Option 1.

III. ANALYTICAL MODEL

To evaluate the performance of the enhanced DHCP, wedefine the evaluation metric: the probability of the success incall setup and handover with the enhanced DHCP. This metricis to estimate the enhancement in the network utilization ofwhite space networks with our proposal.

A. Network Model and Evaluations Procedures

We assume that wireless devices connect to cellular or WiFiTM networks to retrieve the information of white space chan-nels permitted for the broadband Internet access as depictedFig. 2 and Fig. 3 . Thus, in performance analysis, the principleof channel access allows only one white space network tooperate on one white space channel, and after all white spacechannels are occupied, newly-joining WSDs have to wait untilany channel is released. With this occupancy policy, one whitespace network per channel, not only the evacuation of WSDsfrom their operating channel proclaimed by primary users butalso the protection of already-activated white space networkcan be effectively supported. Thus, we adopt the M/M/n/nqueuing model as the evaluation model for our proposal sincethe M/M/n/n queuing model well matches with the occupancypolicy. To evaluate the performance of our proposal, we definethree input streams: one stream is for any arbitrary new andhandover calls, one stream is for handover calls with theassistant of the RRM server or the enhanced DHCP, and thelast one is for remaining calls that experienced a handoverfailure. T1, T2 and T3 are the timeout parameters for the 1st

input stream, the 2nd input stream, and the 3rd input stream

λ1, Τ1

λ2, Τ2 λ4

123

n

.

.

.

Wireless Channelsλ: Arrival RateΤ: Timeout

λ3, Τ3

Fig. 4. Queuing model for the evaluation of our proposal

respectively [7] . We evaluate the performance of our proposalwith the following procedures:

• First, with the queuing model as depicted in Fig. 4, weestimate the utilization of white space networks withconventional handover from cellular or WiFi TM networksto white space networks.

• Second, we estimate the utilization of white space net-works with the handover of our proposal.

Through two evaluation procedures, we analyze the extendto which our proposal decreases the setup time with theassistance of the enhanced DHCP.

We define Tsa to stand for the call setup time of arbitrarynew or handover calls, Thr to stand for the call setup andhandover times with the assistance of the RRM server, Thd tostand for the call setup and handover times with the enhancedDHCP server, and Tcr to stand for the spent time on thehandover that was failed. Based on the previous definitions, wecan assume Tsa < Thd < Thr < Tcr. For the first evaluation,we set T1 to Tsa, T2 to Thr and T3 to Tcr, and for the secondevaluation, we set T1 to Tsa, T2 to Thd and T3 to Tcr. We use(1) to estimate the probability of timeout of the input stream.

W i(t) =

W iE(t)−

3∑k=i

νkWiE(Tk)

1−3∑k=1

νkWiE(Tk)

, Ti−1 < t ≤ Ti, (1)

where W iE(t) = n−λ

n−ΛiE2,nθi−1e

−(n−Λi)t is the ordinaryErlang waiting distribution of the ith input stream, n is thetotal number of open white space channels, and E2,n is Erlangsecond formula. The parameters νi, θi, and Λi are expressedas

Λi =3∑k=i

λk, θi =i∑

k=1

λkTk, νi =λi

n− Λi+1, θ0 = 0, Λ4 = 0,

(2)where λ is the total arrival rate, which is expressed byλ =

∑3k=1 λk = Λ1. With (2), as (2.4) of [7], we use

Pto = Πi(t) =∫ t

0W i(t) dt, i = 1, 2, 3 to calculate the

probability of timeout for input stream 1, 2 and 3. Thus, wedefine λ4 as the new arrival rate that is the sum of three arrivals

810

with timeouts, which is expressed as

λ4 =3∑k=1

λi(1−Πi(Ti)). (3)

With a calculated λ4, we estimate the probability of successin all types of new and handover calls using the followingequation.

Ps = 1− (ρ′)n/n!

1 +n∑k=1

(ρ′)k/k!

, ρ′

= λ4/µ, (4)

where µ is the service rate, and ρ′(= λ4/µ) is a revised traffic

intensity. λ4 for the first evaluation, which aims to analyzethe performance of the network consisting of conventionalnew calls with handover calls with the assistance of the RRMserver in a conventional way, is expressed by λF4 = λ1(1 −Π1(Tsa))+λ2(1−Π2(Thr))+λ3(1−Π3(Tcr)). λ4 for the sec-ond evaluation, which aims to analyze the performance of thenetwork consisting of conventional new calls with handovercalls with the assistance of the enhanced DHCP, is expressedby λS4 = λ1(1−Π1(Tsa))+λ2(1−Π2(Thd)+λ3(1−Π3(Tcr)).Thus, we calculate Ps for the first evaluation with λF4 while wedo Ps for the second evaluation with λS4 . Before conductingthe previous evaluation procedures, we need to figure out howmuch time we can save from the setup time with our proposal.

B. Setup Time of Network Selection

In our proposal, the DHCP clients conduct the last steps, theconnection to the external resource management server and theretrieval of a list of connectable access networks, between theDHCP Discover and the DHCP Offer or between the DHCPRequest and the DHCP Ack. Let denote Tα the setup time ofthe link layer of WiFi TM or cellular network and denote Tβthe setup time of the link layer of alternative access network,for instance, white space network. In addition, let denote Tγthe elapsed time between the DHCP Discover and the DHCPAck and denote Tδ the time spent for the retrieval of a list ofconnectable access networks.

The total time of the call setup and handover with theRRM server in a conventional way is expressed as Thr =Tα +Tβ + 2×Tγ +Tδ . The conventional mechanism requirestwo times of the DHCP initial operation while our proposaldoes one and half or one time of it for the option 1 and theoption 2 respectively. Thus, the total of the hand-over with ourproposal is expressed as Thd = Tα+Tβ+φ×Tγ+ψ×Tδ , whereφ is 1.5 or 1, and ψ is the constant value ( ≤ 1) that calculatesa reduced time as a result of merging the retrieval of the list ofthe ANM. If the DHCP server 1 once retrieved the ANM of thelocation, it can present the retrieved ANM without connectingto the external resource management server. Thus, in this case,ψ is zero. In performance analysis, we ignore the impact ofincrease in packet loss because of increase in the packet sizeto contain proposed options in DHCP messages. According to[8], the packet size in user datagram protocol (UDP), which isa transmission control protocol for DHCP, has the minimum

TABLE IPARAMETERS FOR EVALUATION OF THE ENHANCED DHCP.

Description Unit ValueN number of channels 30λ Arrival Rate [1/17-1/50]k min session duration time sec 120p max session duration time sec 2400E(Td) mean session duration time sec 333µ session termination rate 1/333Tα L2 setup time of WiFi TM sec 2.5Tβ L2 setup time of White Space sec 3.5Tγ DHCP setup time sec 1Tδ A retrieval time from the RRM sec 2

impact on the loss rate such as 0.2 or 0.1% with differentpacket size. Thus, we assume that increase in the DHCPmessages has no significant impact on increase in the packetloss rate. The difference of the setup time, Tdf = Thr − Thd,has a maximum value when ψ is zero while it has a minimumvalue when ψ is one. Thus, for the option 1, the maximumdifference is 0.5 × Tγ + Tδ , and the minimum difference is0.5× Tγ . For the option 2, the maximum is Tγ + Tδ , and theminimum is Tγ .

IV. NUMERICAL RESULTS

Based on latency budget of WiFiTM systems defined in[9], we define the values of parameters in Table I for thenumerical evaluation. The probability density function (pdf)of the service duration time follows the bounded Paretodistribution, which is expressed by

E(Td) =kω

(1− (k/p)ω

)

(ϕ

ω − 1

) (1

kω−1− 1

pω−1

), (5)

where ω is 1.2 [10]. With k and p in Table I, we haveE(Tarr) = 333 seconds, and µ is calculated by 1

E(Tarr) =

1/333. From [11], the duration time of speech of 120 sec-onds is selected as the minimum session duration time andfrom [12], the time of 192 (the average time to watch oneweb video) × 12 seconds is selected as the maximum time.The layer 2(L2) setup of WiFi TM consists of 802.11 scan,association and authentication procedures, and from [9], wehave Tα = 2.5 seconds that is approximate to the sum of theworst case. We also have Tβ = 3.5 seconds for the L2 setupof white space devices, and it is one second greater than theL2 setup of WiFi TM because of a burden of scanning morechannels.

A. Decrease in Setup Time

With given Tα = 2.5, Tβ = 3.5, Tγ = 1, and Tδ = 2, asthe DHCP server can store the retrieved ANM in their localmemory or storage, which means ψ reaches zero, Tdf increasesup to 3 seconds or 2.5 seconds. The update between the DHCPserver and the RRM server in accordance with the update ofthe ANM causes increase in Tdf . However, the frequent updatebecomes a burden of network traffic between the DHCP serverand the RRM server. In this case, since ψ nears one, Tdfdecreases to 1 second or 0.5 second. In the selection of an

811

0.0

0.1

0.2

0.3

The

Pro

babi

lity

of T

imeo

uts,

Pto

Pto of Π2 with Thr = 10

Pto of Π2 with Thd= 7.5

Pto of Π1 with Tsa = 4.5 and Thr= 10

Pto of Π1 with Tsa = 4.5 and Thd= 7.5

Arrival Rate, λ0.0210.0240.0320.0380.058 0.047

Fig. 5. The calculated Pto with λ1 = λ2 = λ3

active access network, the DHCP clients with our proposalcan expect the fast entry as long as ψ is less than one.

In fact, ψ is not greater than one since the extendedDHCP server launches the update of the ANM in accordancewith the activation of the new clients. If the ANM onlyprovides a list of the available access networks, the list willnot be so frequently updated until new access networks areadded. In addition, even if the ANM also contains dynamicchannel information of wireless environments, the ANM willbe updated in proportional to the number of changes occurredin wireless environments of access networks. For instance,let us assume that the RRM server manages a list of whitespace with the occupancy status that are reported by associatedWSDs [13]. As more WSDs are activated, more changes ina list of unoccupied white space channels occur to matchthe changed channel occupancy status, which results in moreupdates in the ANM. However, sometimes, the activation ofnew clients will not change the ANM if the clients decidednot to operate on white space. Thus, we regard that the oneupdate of the ANM between the DHCP server and the RRMserver is equal to or less than the one retrieval of the ANMfrom the RRM server. That is a reason why ψ is equal to orless than one.

B. Probability of the Success of Call Setup

We set Tsa to 4.5 = Tβ + Tγ with the assumption thatwireless devices have the cache about a list of available whitespace channels of a current location, which means that thedevices can directly execute the layer 2 setup of white spacenetworks. We set Thr to 10 = Tα + 2× Tγ + Tβ + Tδ as thenormal call setup and handover with the RRM server, and weset Thd to 7.5 = Tα + 1.5 × Tγ + Tδ . The difference, 2.5,between Thr and Thd refers to Tdf that is described at theprevious section. For Tcr, we set it to 12 = Thr + 2 as thetime in the failure of the handover trial. For the simplicity ofanalysis, we set λ1, λ2, and λ3 to the same value. At the firstevaluation, we have three inputs, normal calls (λ1, Tsa), callsetup and handover with the RRM (λ2, Thr), and calls residingafter the handover is failed (λ3, Tcr) while at the second

20

22

24

26

28

30

' with Tsa + Thd

' with Tsa + Thr

New

Tra

ffic

Inte

nsity

, '

Arrival Rate,

(a) 1.12

0.058 0.047 0.038 0.032 0.024 0.021

' < 20

(a) The calculated ρ′

0.85

0.90

0.95

1.00

(a) 0.028

Arrival Rate, 0.0210.0240.0320.0380.058 0.047

The

Prob

abili

ty o

f suc

cess

in C

hann

el A

ssig

nmen

t), P

s

Ps with Tsa + Thr Ps with Tsa + Thd

(b) The calculated Ps

Fig. 6. Probability of the success in handover calls with λ1 = λ2 = λ3

evaluation, we have (λ1, Tsa), (λ2, Thd), and (λ3, Tcr). Weobserve that Pto = Π2 of (λ2, Thr) is greater than Pto = Π2 of(λ2, Thd) as depicted in Fig 5. At λ1 = λ2 = λ3 = 0.058, thedifference is over 0.05, and the difference decreases as λ1, λ2,and λ3 decrease. Those observations indicate that the handoverwith the enhanced DHCP has a less probability of timeoutcompared to the probability of timeout for the conventionalhandover.

Since our proposal contributes to decrease in the numberof new arrivals to be expired, it results in increasing ρ

′that

is calculated with λ4 as depicted in Fig 6(a). The (a) label inFig 6(a) shows that the difference of traffic intensity betweenρ

′of λS4 and ρ

′of λF4 is about 1.12. This means that among

n = 30, our proposal can accommodate one or more WSDscompared to the conventional approach. As depicted in the (a)label in Fig 6(b), with our proposal, Ps of one wireless deviceis 0.028 less than Ps of another device with the conventionalapproach since the utilization with our approach reaches thelimit, n = 30 earlier than the utilization of the conventionalapproach. With asymmetric arrival rates of λ1, λ2, and λ3,which λ2 is two times greater than λ1 and λ3, we observe that

812

0

1

2

3

4

0.9

0.7

0.5

0.3

0.1

0.90.7

0.50.3

0.1

Del

ay (s

econ

ds)

Miss rate

, P mProbability of failure in

the retrieval, Pf

Delay < 2 seconds

Fig. 7. Calculated delay in the retrieval of the spectrum map

the difference between ρ′

of λS4 and ρ′

of λF4 is about 0.5 atλ1 = λ3 = 0.033 and λ2 = 0.066. In addition, when λ2 is thehalf of λ1 such as λ1 = 0.07 and λ2 = 0.035, the differenceis about 0.7. We observe the similar results that ρ

′with our

approach reaches the limit (n = 30) more closely than ρ′

withthe conventional approach with the most cases of asymmetricarrival rates. Previous observations indicate that our proposalwith the enhanced DHCP can increase the network utilizationcompared to the conventional approach with the RRM server.

C. Delay in the Internet access

As a drawback of our proposal, delay in the IP assignmentof the WiFiTM or 3G network can cause the DHCP clients toexperience delay in the Internet access. Through maintainingpreviously retrieved spectrum maps, the DHCP servers candecrease this delay. Let denote D the delay in the retrieval ofthe spectrum map from the RRM server at the DHCP server,which is expressed by

D = d0 + (Pm × d1)× (1− Pf ) + Pf × d2, (6)

where d0 is the internal delay of the DHCP server for theIP assignment, Pm is the miss rate in finding the spectrummap among the stored spectrum maps, and d1 is the delaytime for retrieving the spectrum map from the RRM server.Pf is the probability of failing to retrieve the spectrum mapfrom the RRM server, and d2 is the delay time in the case ofthe retrieval of the spectrum map failed. For the conventionalapproach, D is d0, and for the simplicity of the analysis, weset d0 to zero. Thus, compared to the conventional approach,the delay penalty is (Pm × d1)× (1−Pf ) + Pf × d2. We setd1 to Tδ = 2 and d2 to the two times of Tδ as the timeout.

Fig. 7 depicts the results of the calculated delay using(6). When Pf is less than or equal to 50%, the calculateddelay is less than Tδ = 2 regardless of what Pm is. WithTdf = 2.5 presented in the previous section, we can expect thatour proposal at least shortens half a second. This observation

indicates that the RRM server can offer the spectrum map witha less delay (≤ Tδ) if Pf and Pm are less than or equal to 50%.As more spectrum maps are internally stored, decrease in themiss rate will result in decrease in the delay of retrieving thespectrum map. This delay can be deteriorated if we have theDHCP server and the RRM server integrated as one server.

V. CONCLUSION

This paper proposes the enhanced DHCP that extendsDHCP messages to include spectrum map. Our proposal aimsto reduce a signaling overhead for handover from the conven-tional access networks to the alternative networks, white spacenetworks. We propose the retrieval of the spectrum map duringthe DHCP operation for the first IP assignment and expect thatthe embedded retrieval of the spectrum map can contribute toincreasing the network utilization. In performance analysis, weconduct numerical evaluation with handover cases from cellu-lar or WiFiTM networks to white space networks. Accordingto the analysis results, as the arrival rate,λ, reaches a certainlevel, which means we have no more unoccupied channels,with the enhanced DHCP, we can accommodate up to oneor more white space devices, and as a result, the networkutilization increases. We expect that we can further use thisfast retrieval for any new ranges of open spectrum for thebroadband Internet access.

REFERENCES

[1] T. Taleb, A. Jamalipour, Y. Nemoto, and N. Kato, “DEMAPS: A Load-Transition-Based Mobility Management Scheme for an Efficient Selectionof MAP in Mobile IPv6 Networks,” IEEE Transaction on VehicularTechnology, vol. 58, no. 2, Feb. 2009.

[2] Patent: Method and system for discovering vacant DTV channels usingDHCP server location Pub.No.: US 2009/0217333 A1

[3] M. Kobayashi, H. Nakayama, N. Ansari, and N. Kato, “Robust andEfficient Stream Delivery for Application Layer Multicasting in Hetero-geneous Networks,” IEEE Transaction on Multimedia, vol. 11, no. 1, Jan.2009.

[4] T. Aura, M. Roe, and S. J. Murdoch, “Securing NetworkLocation Awareness with Authenticated DHCP,” [online]http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.127.2095

[5] M. Buddhikot, A. Hari, K. Singh, and S. Miller, “MobileNAT: A NewTechnique for Mobility Across Heterogeneous Address Spaces,” [online]kundansingh.com/papers/mobilenat-monet05.pdf

[6] A. Dutta, S. Das, D. Famolari, Y.Ohba, K. Tani-uchi, V.Fajardo, R. M. Lopez, T.Kodama, and H. Schulzrinne, “Seamlessproactive handover across heterogeneous access networks,” Wireless PersCommun , DOI 10.1007/s11277-007-9266-3

[7] R. B . Haugen and E. Skogan, “Queueing systems with stochastic timeout,” IEEE Trans. Communications, COM-28, No. 12, 1980.

[8] Y. A. Wang, C. Huang, J. Li, and K. W. Ross, “Queen: Estimating PacketLoss Rate between Arbitrary Internet Hosts,” The tenth Passive and ActiveMeasurement conference, March 2009

[9] A. Alimian and B. Aboba,“Analysis of roaming techniques,”IEEE 80211-04/0377r1, Cited October 19th 2004 fromhttp://wwwieee802org/11/Documents/

[10] M. C. Weigle, “Improving Confidence in Network Simualtions,” Pro-ceedings of the 2006 Winter Simulation Conference, 2006.

[11] L. Correia, Mobile Broadband Multimedia Networks:techniques, modelsand tools for 4G Academic Press, 2006.

[12] P. Gill, M. Arlitt, Z. Li, and A. Mahanti, “YouTube Traffic Characteri-zation: A View From the Edge,” Internet Measurement Conference (IMC2007). Nov. 2007.

[13] S. Yoon, K. Lim, and J. Kim, “Cross-Layer Dynamic Spectrum MapManagement Framework for White Space Applications,” EURASIP Jour-nal on Wireless Communications and Networking, vol. 2010, Article ID870976, 2010.

813