1

Idle Mode for Deep Power Save in IEEE 802.11 WLANsSunggeun Jin, Kwanghun Han, and Sunghyun Choi

Abstract: Along with the wide acceptance of IEEE 802.11 WirelessLocal Area Network (WLAN), new applications such as InternetProtocol (IP) telephony over WLAN are fast emerging today. Forbattery-powered IP phone devices, the life time extension is a keyconcern for the market acceptance while today’s 802.11 is not opti-mized for such an operation. In this paper, we propose a novel IdleMode operation, which comprises paging, idle handoff, and de-layed handoff. Under the idle mode operation, a Mobile Host (MH)without any active session does not need to perform handoff withina predefined Paging Area (PA). Only when it enters a new PA, anidle handoff is performed. The proposed idle mode allows an MHwithout traffic to extend its life time. We develop a new analyti-cal model in order to comparatively evaluate our proposed scheme.The numerical results demonstrate that the proposed scheme out-performs the existing schemes with respect to power consumption.

Index Terms: IEEE 802.11 WLANs, idle mode, power saving.

I. INTRODUCTION

Recently, IEEE 802.11 Wireless Local Area Network(WLAN) became a prevailing technology for the broadbandwireless Internet access. Along with that, new types of appli-cations such as Internet Protocol (IP) telephony over WLAN arefast emerging today. For the wide market acceptance of battery-powered IP phones in the growing 802.11 WLANs, IP phone’spower consumption efficiency appears a key concern for the mo-bility management in the 802.11 WLAN while today’s 802.11 isnot optimized for this. The reason why the 802.11 WLAN pro-vides poor efficiency in power consumption for the IP phones isrooted in the fact that the 802.11 WLAN Medium Access Con-trol (MAC) protocol [1] defines only two operational modes,which an Mobile Host (MH) can operate in, namely, ActiveMode (AM) and Power Save Mode (PSM).

In both modes, an MH has to always remain associated withan Access Point (AP) even when there is no traffic to/from it.This issues a critical problem that it has to perform handoffsat every AP cell boundary in order to maintain the associationwith an AP. The inevitable handoffs, at every AP cell boundary,cause MH without active traffic1 to waste precious battery powerin vain. As the handoff frequency, approximately proportionalto the MH’s speed, increases, the MH consumes more power.

Even worse, when IEEE 802.11i [2] is employed for secu-rity enhancement, a larger amount of message exchanges during

An earlier version of this paper was presented in Proc. IEEE ICC’06, June2006 [27].

Sunggeun Jin is with ETRI, Daejeon 305-700, Korea.Kwanghun Han and Sunghyun Choi are with School of Electrical En-

gineering and INMC, Seoul National University, Seoul 151-744, Korea(email:schoi@snu.ac.kr).

This work was supported by the IT R&D program of MKE/KEIT. [2009-F-044-02, Development of cooperative operation profiles in multicell wireless sys-tems]

1We refer to an MH without active traffic as an idle MH.

the handoff operation are expected, and these incur more powerconsumption. Currently, IEEE 802.11r [4] is being developedin order to overcome the overhead for performing the 802.11i-related security operations during the handoff. Despite of theefforts, 802.11r does not remove the handoff of idle MH itselfso that it still incurs the waste of power.

That is, IEEE 802.11 WLAN is naturally lack of an efficientsupport of the mobility with respect to power consumption whenthere is no traffic to be served for the MHs. Therefore, it is de-sired to have a new mode of operation, called Idle Mode (IM),on top of the currently-available AM and PSM. Due to the ab-sence of such an IM operation, the combination of the IP pagingand the PSM, called the PSM with IP paging in this paper, havebeen proposed as an alternative to the IM operation [21] thoughthe original aim of the IP paging is to facilitate the integration ofdifferent wireless technologies and the IP paging is independentof Layer-2 (L2) technologies [9].

However, as discussed further later, IP paging is found to beharmful to the power consumption efficiency since the powerconsumption under the IP paging increases as the amount ofbroadcast/multicast traffic in the network increases while thereare frequent broadcast/multicast frame transmissions in typicalWLANs. This fact implies that an MH should wake up fre-quently than common expectations in order to manage broad-cast/multicast frames, and hence, it wastes more power in vain.

In order to overcome the discussed problems, we proposean IM operation, comprising paging, idle handoff, and delayedhandoff, which can be used when an IEEE 802.11 WLANstandard-based MH does not have traffic or on-going sessions.Using the proposed IM operation, the MH can stay in the dozestate requiring very little power. In our scheme, an MH doesnot perform any handoff within a predefined Paging Area (PA).The handoff with the minimum operation, called idle handoff, isperformed only when an MH leaves a PA. The paging providesa way to inform an MH in the IM of a new frame arrival. TheIP-level handoff should be deferred until a paging success in or-der to reduce redundant operations, and hence, it is referred toas delayed handoff.

In [11], the authors also discuss a paging scheme similar toour IM operation for 802.11 WLAN. In their scheme, Track-ing Agent keeps a cache containing MAC addresses and IP ad-dresses for both MH and its associated AP. However, the cacheis updated at every reassociation, i.e., L2 handoff, and hence, itincurs redundant signaling cost and power consumption.

The rest of this paper is organized as follows. In Section II, wediscuss the limitations of the PSM with IP paging. In Section III,we introduce the IM for the IEEE 802.11 WLAN. Additionally,we propose new protocols constituting the IM. In Section IV,we develop an analytical model to evaluate the proposal in termsof power consumption. In Section V, through our mathematicalmodel, we evaluate our proposal and demonstrate the superiorityof our proposal compared with the current PSM with IP paging.

1229-2370/03/$10.00 c© 2003 KICS

2

Table 1. Summary of Acronyms

AM Active ModeAP Access PointBSS Basic Service SetFA Foreign AgentHA Home AgentHD Handheld DeviceIM Idle ModeIP Internet ProtocolL2 Layer-2L3 Layer-3MAC Media Access ControlMH Mobile HostMRA-AP Most Recently Associated APPA Paging AreaPAID Paging Area IDentification numberPSM Power Saving ModePTK Pairwise Transient KeyRRB Remote Request BrokerVoIP Voice over IPVoWLAN Voice over WLANWNIC Wireless Network Interface CardWLAN Wireless Local Area Network

Finally, in Section VI, we conclude this paper with the summaryof our efforts and results. The acronyms used in this paper aresummarized in Table 1.

II. CURRENT LIMITATIONS

Paging is developed to locate an idle MH when there arrivesan incoming call destined to the idle MH. If an access net-work capable of supporting IP, however, does not offer pagingscheme, IP paging can be used as an alternative for the wirelessnetwork level paging. For this reason, researchers have con-sidered that the PSM together with IP paging could be the al-ternative scheme for the 802.11 WLANs [21]. However, sincethe PSM was developed without consideration of IP paging, theuse of IP paging along with the PSM could be an inefficientapproach. In this section, we discuss the reason why the combi-nation of the PSM and IP paging is not suitable as an alternativeto the IM for IEEE 802.11 WLAN.

A. Limitations of the PSM with IP paging

The 802.11 standard specifies that MH’s Wireless NetworkInterface Card (WNIC) can be in either of awake and dozestates [1]. In awake state, it can transmit, receive or sense thephysical channel while it actually continues to sense the chan-nel unless it either transmits or receives a frame. On the otherhand, in doze state, it is not able to transmit nor receive, andhence, consumes very little power. How WNIC switches be-tween these two states is determined by its power managementmode, i.e., the AM and the PSM. A WNIC in the AM alwayskeeps operating in the awake state while the WNIC in the PSMcan change its state between the awake and doze states depend-ing on the traffic pattern. Based on these features, we summarizethe limitation of the PSM with IP paging for an alternative to theIM as follows.

Since an MH running in the PSM stays associated with anAP, handoff is performed at every AP cell boundary in orderto maintain its association, thus resulting in redundant powerconsumption. During a handoff procedure, the MH has to stay

Table 2. Broadcast/Multicast Frame Inter-Arrival Time Statistics

0∼10 ms 10∼100 ms 100ms∼1 s ≥1 s59.73 % 10.50 % 29.41 % 0.36 %

Table 3. Various types of multicast and broadcast frames.

Protocol % Protocol %ARP-Request 67.27 UDP (SSDP) 9.56

UDP (LLMNR) 5.33 IGMP 4.92UDP (WSD) 4.60 UDP (unknown) 3.58

UDP (802.11 IAPP) 2.74 DNS 1.15

in the AM since the handoff can be severely delayed otherwise.In order to employ the PSM with IP paging, several Foreign

Agents (FAs) are grouped to cover an IP paging area. Wheneveran MH crosses an IP paging area, it should perform a locationupdate procedure including an L3 handoff. As proved in [22],it takes several seconds to perform an L3 handoff due to the L3operation features. For this reason, higher mobility drives powersaving schemes to require more power. When a new call arrives,a selected agent broadcasts IP paging message(s) through theentire IP paging area [9]. An MH in the PSM is informed thatit is paged by receiving the broadcast IP paging message, whichall APs in the IP paging area forward after Delivery Traffic In-dication Message (DTIM) transmission.2

However, this policy requires unnecessary manipulations ofbroadcast/multicast frames. The PSM with IP paging schemeis employed when there is no active traffic for an MH. As ex-plained early, IP paging depends on broadcast IP frames to in-form idle MHs that new calls destined to the MHs are about to beestablished. Therefore, an idle MH in the PSM should receivebasically all the broadcast/multicast frames in order to detect theexistence of newly-arriving calls although most frames are ac-tually useless, and then, the received broadcast/multicast framesare forwarded from MAC to IP layer.

As to practical devices, an MH consists of WNIC and Hand-held Device (HD) where WNIC is attached. HDs are portableequipments such as Personal Digital Assistance (PDA) or smartphone. IEEE 802.11 WLAN standard-based MAC layer residesin a WNIC while IP layer, as a part of Operating System (OS),is embedded in HDs. It implies that HD needs to process theseforwarded IP frames by consuming considerable energy. In or-der to verify the reasoning, we measure the inter-arrival timesof broadcast/multicast frames, i.e., the times between two con-secutive broadcast or multicast frames, for an hour in NESPOT,which is a large-scale commercial WLAN operated by KoreaTelecom (KT). Additionally, we obtain another statistics in or-der to take a closer look at what kinds of broadcast/multicastframes are being transmitted in the network. For the measure-ment, we have used three MHs associated with an AP in theNESPOT.

Tables 2 and 3 show the inter-arrival times and the types ofthe measured broadcast/multicast traffic. From the tables, sur-prisingly the inter-arrival times under 10 ms represent the majorportion. The average frame inter-arrival time is 116.7845 ms.This fact shows that there will be a broadcast/multicast frameevery beacon interval (assuming 100 ms beacon interval) in av-

2DTIM transmission interval is a count of the number of beacon frames, ofwhich transmission period is typically 100 ms.

3

erage, and hence, the MH has to wake up often, e.g., every bea-con interval, to receive these frames. For this reason, any MHsadopting an IP paging are compelled to consume their energy invain in order to receive useless frames.

We classify the collected packets by referring to the destina-tion port since the destination port typically indicates the us-age of the frame [13]. However, we cannot recognize the usagewhen a multicast frame contains a particular port number, ofwhich usage is unknown. Table 3 shows what kind of multi-cast and broadcast frames are in the wireless network. The ratiois obtained by dividing the number of the corresponding typeframes by the total collected broadcast/multicast frames.

In this table, we observe that ARP-Request frames occupy67.27 % of the broadcast/multicast frames. The ARP-Requestframes are necessary to IP address management for the Dy-namic Host Configuration Protocol (DHCP). When there is noresponse corresponding to an ARP-Request, DHCP server de-termines that it should withdraw the corresponding IP addressallocated to an MH. UDP (SSDP) accounting for 9.56 % repre-sents UDP multicast packets for Simple Service Discovery Pro-tocol (SSDP), which is designed for Universal plug-and-play byMicrosoft and Hewlett-Packard. Internet Group ManagementProtocol (IGMP) and UDP (LLMNR) are utilized for the mem-bership management of Internet Protocol multicast groups andLink Local Multicast Name Resolution protocol, respectively.UDP (WSD) is used for Web Service Discovery (WSD) proto-col. UDP (unknown) indicates the broadcast/multicast framesof which usage is not known. We do not present the rest frames,of which portion is less than 1 %.

Moreover, IPTV services relying on IP broadcast/multicasttransmissions are launched for wired network in Korea. Natu-rally, the IPTV services are expected to be served in wirelessnetworks soon. It implies that we will face more congested IPbroadcast/multicast traffic in wireless network in the near fu-ture. Lastly, the original IP paging is targeted at the integra-tion of heterogeneous wireless networks by providing pagingscheme in IP layer. However, under wireless network exploitingits own MAC-specific paging scheme, IP paging actually wouldprovokes redundant operations since both MAC and IP providethe same functionality, i.e., paging. Nevertheless, due to theabsence of a paging in IEEE 802.11 WLANs, IP paging wouldhave been regarded as a useful candidate for the MAC-level pag-ing [21].

III. PROPOSED IDLE MODE OPERATION

A. Definition of Idle Mode

In order to overcome the problems discussed in Section II, wedefine a new mode, i.e., IM, for IEEE 802.11 WLAN. By utiliz-ing the IM, we allow idle MHs to minimize required operations,thus leading to high power saving efficiency. When an MH is inthe IM, it performs only essential operations to wake up to catcha newly-arriving call in time. The necessary operations for theIM are defined as follows:1. A handoff does not occur at every cell boundary unlike anMH in the PSM. A handoff, called idle handoff, is performedonly when an MH leaves a PA to enter another PA.2. When an MH is in the IM, the MH is not associated with anyAP. The only thing that the MH in the IM has to do is to listen

Home-APMH

IdleMode-Request

IdleMode-Response

MRA-AP

Authentication Request

Authentication Response

Page-Notify

response

Tracking beacon procedure & AP reselection

Packet arrives (triggering paging)

APs in PA

Tracking beacon procedure & AP reselection

Page-Notify

response

Tracking beacon procedure & AP reselection

Paging Success

Reassociation Request

Reassociation Response

Fig. 1. Procedure for the idle mode operation.

to the beacons periodically at every predefined interval in orderto switch itself to the Active Mode when a frame destined toitself arrives. The typical beacon listening interval for receivingbeacons to wake up is set to 1 s, while beacons are transmitted byAPs every 100 ms typically. It should be noted that the call setuplatency increases proportional to the beacon listening interval.3. Only a successful paging makes an MH in the IM enter theActive Mode.

Security is indispensable for proper VoWLAN services. Theemerging 802.11r standard is expected to provide efficient secu-rity schemes for fast roaming when a MH moves across APs [4].We integrate the IM operation with those schemes defined in the802.11r.

B. Protocols for Idle Mode

A number of neighboring AP cells are grouped into a PA.The APs belonging to different routers can also be grouped intoa single PA. The APs in the same PA have the same identifier,which is broadcast through the beacons with a newly-definedPaging Area Identifier (PAID) field. Each MH in the IM candifferentiate a PA via its PAID.

We define a new procedure in order to support the IM. Fig. 1shows the procedure when an MH enters and leaves the IM. Af-ter a session (e.g., a VoIP session) completion, the MH transmitsan IdleMode-Request frame, a management frame, to enter theIM. After receiving the corresponding IdleMode-Response fromthe AP, the MH in the IM can move around within the same PAwhile the AP, which transmitted the IdleMode-Response, main-tains the information regarding the MH in the IM. The informa-tion is used to support the movements to other PAs, security, andcall setup process in the future.

This AP is referred to as Home-AP. After entering the IdleMode, the MH starts listening to beacons periodically (e.g., ev-ery one second). Even when an MH recognizes the change ofAP cell through the beacon information, the MH only keepslistening to the beacons as long as the MH stays in the samePA. This continuous beacon listening operation is called AP-

4

Tracking beacon procedure

& AP reselection

Update PA

Fig. 2. Idle handoff.

reselection. For an efficient AP-reselection, there could be manyoptimization issues as addressed in [3, 7, 8]. However, we donot consider the AP-reselection issues since they are beyond thescope of this paper.3

When a frame destined to a particular MH in the IM arrives atthe Home-AP, the Home-AP broadcasts a Page-Notify messageto all the APs, belonging to the same PA, which in turn startpaging the destination MH. That is, the APs convey the paginginformation via their beacon frames. If an MH recognizes that itis paged by receiving such beacon(s) from an AP, it attempts toassociate with the AP by transmitting a Reassociation-Requestframe. After finishing all the preparations for serving the MH,the new AP replies to the MH with a Reassociation-Responseframe and broadcasts Paging-Success to the APs in the same PAto stop paging operations of these APs. Along with that, aftera successful paging, the MH begins a delayed handoff opera-tion as presented below. Ultimately, the proposed IM enablesthe 802.11 WLAN to page the deeply sleeping MHs only in L2without depending on the IP paging since the PSM with IP pag-ing has redundancies in paging as explained earlier.

The paging may fail when a new call arrives in case thatan MH in the IM does not update its movement to a newly-entered PA. However, the paging failure probability is ignorableas proved in Appendix I. In spite of the ignorable paging fail-ure probability, it is required to cope with the paging failure forreliable service. In case of the paging failure, the paging retryshould be conducted T bli later since an MH in the IM requiresT bli to recognize its movement to a new PA.

C. Idle and Delayed Handoff

Idle handoff is a handoff that is performed whenever an MHin the IM moves across a PA boundary. Fig. 2 shows the proce-dure for idle handoff. After an MH enters a new PA, which canbe identified by a newly-received beacon, it transmits an Idle-Handoff-Request frame including the Basic Service Set IDenti-fier (BSSID) of its Home-AP. If it fails, it can retry the transmis-sion. Upon receiving the Idle-Handoff-Request frame, the newAP receiving the frame forwards it to the Home-AP.4 In casethat a Home-AP receives an Idle-Handoff-Request frame, it up-

3For example, the Neighbor Report information from the emerging802.11k [3] will make this possible.

4The mapping between an AP’s MAC address and its IP address can be ob-tained from a server as defined in IEEE 802.11F [5]. Note that the 802.11F is nota standard. The mapping information can be utilized when a new AP forwardsthe Idle-Handoff-Request frame to the Home-AP.

New FA AP

Paging Success

MH

Router-advertisement broadcasting

Home-AP

Buffered frame forwarding

HA

Router-advertisement broadcasting

Binding update

Buffered frame forwarding

Fig. 3. Delayed handoff.

dates the PAID for an idle MH with the value conveyed throughIdle-Handoff-Request. After transmitting Idle-Handoff-Request,the MH resumes listening to the beacons periodically in order toreceive the paging information. The AP, which is involved withthe 2-way frame exchange, is referred to as Most Recently Asso-ciated AP (MRA-AP).

When there is at least one idle handoff, a Home-AP transmitsa Page-Notify to an MRA-AP, which in turn forwards it to all theAPs in the same PA. Since our proposed scheme enables 802.11WLAN to keep track of the locations of the MHs in the IM, IP-layer related operations including IP paging becomes redundant.That is, our proposed protocol replaces IP paging. Therefore, inour approach, handoff operations related to an IP layer are post-poned until a successful completion of the paging even when anMH needs to conduct IP-layer handoff by departing the cover-age of an FA, where a Home-AP is attached. For this reason, werefer to this handoff operation, which delays the activation of IPlayer, as delayed handoff.

We define a protocol for delayed handoff as shown in Fig. 3.During the paging operation, frames destined to an idle MH arebuffered at the Home-AP. After a successful paging, the Home-AP forwards the buffered frames to a new AP, which the pagedMH is newly associated with. While the Home-AP forwardsthose frames, the MH begins listening to Router-Advertisement-Messages, which an FA broadcasts to provide the IP-layer hand-off. By receiving Router-Advertisement-Messages, the MH caninitiate an IP handoff deferred until the paging.

However, it incurs a corresponding call setup latency. There-fore, we provide a remedy reducing the latency by utilizingProxy MIP [6]. Additionally, we prove that there exists a trade-off relationship between the power consumption, which is rep-resented by a signaling cost, and the call setup latency. Theprevious study only deals with IP layer procedure, and hence,we provide an integrated scheme combining the proposed IMand the IP layer procedure as follows:

We design FAs to perform an MIP handoff on behalf of an idle

5

Home Agent

Proxy MIP procedure Proxy MIP procedure

FA FA FA FA FA

Proxy MIP procedure

Fig. 4. Proxy MIP procedure.

MH, which is referred to as PMIP procedure, whenever an idleMH enters an FA’s coverage. An idle MH neither performs anMIP handoff nor receives broadcast/multicast IP packets, thusreducing power consumption as well as signaling cost as well.In fact, FAs, however, do not need to conduct a PMIP proce-dure whenever an idle MH enters their coverage since an idleMH does not have a session. We assume FAs perform a PMIPprocedure only if an idle MH reaches the coverage of a new FAlocated at a predefined distance from the old FA (oFA), whichhad performed the latest PMIP procedure as shown in Fig. 4.

Keeping these features in mind, we develop a protocol for the802.11 WLANs as follows: (1) an MH enters the IM by com-pleting an idle mode entrance procedure involved with a Home-AP. The Home-AP begins managing the location of the MH af-ter activating PMIP in the FA connected with the Home-AP. (2)When the idle MH performing the idle handoff moves to thecoverage of a new PA, the corresponding MRA-AP determineswhether the idle MH enters a coverage of a new FA by queryingthe Home-AP about the previous FA, where the MH stayed.

In this case, the MRA-AP5 obtains the distance between thenew FA and the oFA. If the distance is equal to a predefinedvalue, it requests the new FA (nFA)6 to conduct the PMIP pro-cedure. Additionally, it requests oFA to remove obsolete contextabout the idle MH and takes over the role of Home-AP. Other-wise, it informs the Home-AP that the MH is in its PA. (3) SinceFAs do not perform the PMIP procedure whenever an idle MHenters a new FA-coverage, IP packets destined for the idle MHarrive at oFA. For this reason, the oFA triggers the Home-AP topage the idle MH. Consequently, the paged idle MH can con-duct an MIP handoff by itself after the completion of a networkentry procedure for the wireless network while oFA forwardsthe arrived IP packets by querying the information about underwhich FA the paged MH is from the Home-AP. As a result, thisproposal enables an idle MH to fully exploit the power savingefficiency, which the idle mode can provide by removing thenecessity to receive broadcast/multicast IP packets.

D. Security for Idle Mode Operation

The IM is designed to work with the existing and emerg-ing 802.11 security schemes including the 802.11i [2] and the802.11r [4]. In the 802.11i, an MH that handoffs to a new AP

5MRA-AP can be designed to have a mapping table containing distancesamong FAs based on pre-configured location information for each FA.

6nFA represents the new FA performing PMIP procedure.

performs the 802.1x authentication procedure prior to the 4-wayhandshake. However, an MH in the IM keeps disassociated froma WLAN. It implies the WLAN does not need to provide it witha secure service. For this reason, an MH employing the IMcan defer the authentication procedure up to the end of a suc-cessful paging, thus incurring less power consumption. Sincethe 802.11i-based commercial networks are already deployed inmany places, we consider the 802.11i WLAN for our evalua-tions.

In an 802.11r WLAN, Fast Transition (FT) key hierarchy isdesigned to accelerate handoff by providing hierarchical keymanagement strategy to establish Pairwise Transient Key (PTK)fast. However, since an idle MH does not have any data to beserved, it does not need to derive the PTK whenever it movesto another AP but computes fresh PTK for secure data trans-mission with a successful paging. Excepting the secure datatransmission, when an idle MH conducts idle handoff, secureIdle-Handoff-Request should be forwarded to Home-AP. In the802.11r, Remote Request Broker (RRB) assumes the role offrame forwarding, and hence, we adopts the functions of RRBin order to support idle handoff. When an MH enters the IM,the idle MH and Home-AP do not release the PTK for futureuse. An idle MH trying to transmit Idle-Handoff-Request frameencrypts the frame with the PTK, which is used for secure trans-mission with Home-AP, and then, transmits it to an MRA-AP.Upon receiving the encrypted Idle-Handoff-Request frame, anRRB in the MRA-AP forwards the frame encrypted with thePTK to Home-AP guaranteeing secure validity.

IV. POWER CONSUMPTION ANALYSIS

The idle mode is developed to improve power consumptionefficiency, and hence, we analyze the performances of both theIM and the the PSM with IP paging in terms of power consump-tion. For the numerical analysis, we separate MH into WNICand HD, which take roles of MAC and IP, respectively sincetheir power consumption is different from each other.

A. Assumptions

We consider VoIP telephony as a target application for ouranalysis. When VoIP phone is used, a session is initiated by anincoming or outgoing call. In order to indicate the operationalstatus of VoIP phone, we define two states for the service asfollows:1. State 1: An MH holds an active session. In other words,WNIC is in the awake state and the HD is powered on. The MHhas an session for traffic. It performs handoff whenever it movesacross AP cells.2. State 2: An MH stays idle without holding active session. Itimplies that WNIC switches between the awake and the dozestates at every predefined time interval periodically in order toreceive beacons including the paging information. If the IM isutilized, WNIC performs idle handoff whenever it leaves a PA.A successful paging or outgoing call makes the MH enter State1.

For general assumption, we do not consider microsleeping7

for on-going VoIP session in State 1, i.e., the state transition

7Microsleeping is a sleeping for silent voice period while MH holds activesession.

6

I

m

...0M

t1

Mt

KM

t1K

Mt

Fig. 5. Timing for area crossings: the term “area” is used to imply an IPpaging area as well as a paging area.

from the awake state to the doze state does not occur during thewhole on-going session time. In order to determine the steadystate probability, we make the following assumptions:1. Incoming and outgoing calls at an MH occurs according to aPoisson process with rates λin and λout, respectively.2. Session holding time th is generally distributed with a prob-ability density function fh(t) and its expectation is 1/λh.3. Idle time tI is exponentially distributed with the expectation1/λI . Note that λI = λin + λout.

B. Area Crossing Probability

IP paging area encompasses multiple routers, each of whichmay include several APs, and hence, it is reasonable that an IPpaging area comprises several PAs, which in turn consists of sev-eral AP cells. For this reason, IP paging area and PA are struc-tured in hierarchal network topology, where MH would movearound by crossing AP cells while possibly crossing them aswell. Power consumption depends on how many areas includ-ing IP paging areas, PAs, or AP cells an MH passes through.

In Fig. 5, let random variable tI represent an idle time. fI(t)is the probability density function (pdf) of tI . Generally, fI(t)can be an arbitrary pdf, but we assume an exponential distri-bution. Random variable tMi represents area sojourn time atthe ith area. fMi(t) and E[tMi ](= 1/λMi) are the pdf and theexpectation of area sojourn time, respectively. We assume thatarea sojourn times, tMi’s, are i.i.d., and follow the same distri-bution. Therefore, we simply denote tMi and fMi(t) as tM andfM (t), respectively. Let tm be the interval between the begin-ning of the IM and the instance when the MH leaves the firstarea while fm(t) indicates its pdf. fM (t) and E[tM ](= 1/λM )are the pdf and the expectation of tM , respectively. F ∗m(s) andF ∗M (s) represent the Laplace transform functions of fm(t) andfM (t), respectively.

We derive the probability Pr(tM < tI) that an MH crosses anarea during idle time by:

Pr(tM < tI) =∫ ∞

0

∫ ti

0

fI(ti)fM (t)dtdti

= λI

∫ ∞

0

e−λIti

∫ ti

0

fM (t)dtdti

= λIF ∗M (λI)λI

= F ∗M (λI).

Similarly, we have the probability that an MH crosses thefirst area by F ∗m(λI). Now, we easily derive the probabilityPr(K = k) that an MH crosses areas k times during idle timefrom the reasoning: (1) the MH crosses the first area with theprobability F ∗m(λI), and thereafter, (2) it crosses k − 1 PAs with

Fig. 6. Types of cells within an area and outside-regions.

the probability (F ∗M (λI))k−1; finally, it stops crossing the PAwith the probability (1− F ∗M (λI)). The equation is obtained bymultiplying those probabilities summarized by:

Pr(K = k) =

{1− F ∗m(λI), k = 0,F ∗m(λI)(F ∗M (λI))k−1(1− F ∗M (λI)), k ≥ 1,

(1)

where k = 0 implies that MH stays within the first area. Weintroduce another derivation in Appendix II.

C. Area Sojourn Time Probability

Since an area, i.e., either an IP paging area or a PA, is com-posed of a number of cells, we can derive fm(t) and fM (t) asfollows. Fig. 6 shows the cell types in an area and the outside-region types at the outside of an area. In this figure, cells areclassified into different types according to the absolute positionin a sector. A sector is defined by a region between adjacentintersecting lines.

The outside-region type is labeled after the type of its adja-cent cell in an area. A layer is defined as a group of cells withthe same depth from the area center. The probability matrixP (k), where element p(k)

(x,y),(x′,y′) of P (k) represents the proba-bility that MH moves from either a type < x, y > cell or a type< x, y > outside-region to a type < x′, y′ > cell or outside-region exactly at the kth random crossing, can be derived by:

P (k) = P · P (k−1), k ≥ 1,

where P = (p(x,y),(x′,y′)) is the transition matrix, andp(x,y),(x′,y′) is the probability that MH departs either a type< x, y > cell or a type < x, y > outside-region to enter a type< x′, y′ > cell or outside-region.

Let the maximum number of the layers in an area be N .If there exists only a single cell in an area, namely, type< 0, 0 > cell, N = 1. The transition matrix P becomes(N(N+1)

2 × N(N+1)2

)matrix. MH can reach type < N, j′ >

outside-region by departing an area only via type< N − 1, j′ >

7

cell. Using the transition matrix, the probability pk,(x,y),(N,j)

that an MH initially resides at a type < x, y > cell, and movesinto a type < N − 1, j > cell at the k − 1st crossing, and then,departs an area at the kth crossing is derived by [25]:

pk,(x,y),(N,j) =

{p(x,y),(N,j), k = 1,p(k)(x,y),(N,j) − p

(k−1)(x,y),(N,j), k > 1,

(2)

where 0 ≤ j < N − 1.Random variable t(k) represents the time that an MH spends

until leaving an area while it visits k cells. If an MH begins theIM in the area, it is denoted by:

t(k) =

(tr +

k∑

i=2

tc

),

where tr is the random variable for the residual service time [19]defined by that required to leave a cell when an MH begins theIM in the area. Let tc denote a random variable for cell sojourntime. E[tc](= 1/λc) is its expectation. If an MH passes throughan area while visiting k cells, t(k) is derived by:

t(k) =k∑

i=1

tc.

We derive F ∗m(s) and F ∗M (s) by considering how many cellsan MH visits until leaving an area. Let f(k)(t) represent the pdfof t(k). Accordingly, fm(t) is derived by [24]:

fm(t) =∞∑

k=1

N−1∑n=0

n−1∑y=0

N−2∑

j=0

ψ(n, y)pk,(n,y),(N,j)f(k)(t), (3)

where ψ(n, y) is the probability that MH starts its IM when stay-ing in a type < n, y > cell. In a hexagonal area, there are sixcells of the same type except for type < 0, 0 > cell. Assum-ing that the total number of cells in an area is S(N), ψ(n, y) isderived by:

ψ(n, y) =

{1

S(N) , n = 0 and y = 0,6× 1

S(N) , otherwise.

In this equation, S(N) is given by 6× N(N−1)2 + 1. Since the

pdf of tr is defined by fr(t) = λc

∫∞tfc(τ)dτ [19], the Laplace

transform function of fr(t) is derived by:

F ∗r (s) =λc

s(1− F ∗c (s)).

From Eq. (3), the Laplace transform F ∗m(s) is derived by:

F ∗m(s) =∞∑

k=1

N−1∑n=0

n−1∑y=0

N−2∑

j=0

ψ(n, y)pk,(n,y),(N,j)

×λc

s(1− F ∗c (s))(F ∗c (s))k−1,

(4)

When an MH enters an area, it begins its movement at type< N − 1, y > cell. It implies that the probability that the MHleaves the area at the kth cell crossing becomes pk,(N−1,y),(N,j).Therefore, the pdf of area sojourn time required to cross an areais derived by [24]:

fM (t) =∞∑

k=1

N−2∑y=0

N−2∑

j=0

ϕ(N−1,y)pk,(N−1,y),(N,j)f(k)(t), (5)

where ϕ(N−1,y) is the probability that MH enters an areathrough a type < N − 1, j > cell at the first crossing. The val-ues of ϕ(N,y) are obtained as follows:

ϕ(N,y) =

{3

2N+1 , y = 0,2

2N+1 , otherwise.

Using Eq. (5), the Laplace transform function F ∗M (s) offM (t) is derived by:

F ∗M (s) =∞∑

k=1

N−2∑y=0

N−2∑

j=0

ϕ(N−1,y)pk,(N−1,y),(N,j)(F ∗c (s))k.

(6)

We apply Eqs. (4) and (6) to the derivations of Pr(KIP = k)and Pr(KPA = k), each of which represents the probability thatan MH crosses IP paging area and PA k times during an idle timetI , respectively. For the purpose, we use new notations F ∗IP (s)andF ∗ip(s) instead ofF ∗M (s) andF ∗m(s) for Pr(KIP = k). Sim-ilarly, F ∗PA(s) and F ∗pa(s) correspond to F ∗M (s) and F ∗m(s) forthe derivation of Pr(KPA = k), respectively. Consequently, wesummarize the corresponding Laplace transform functions asfollows:

F ∗IP (s) =∞∑

k=1

L−2∑y=0

L−2∑

j=0

ϕ(L−1,y)pk,(L−1,y),(L,j)(F ∗c (s))k

∣∣∣∣s=λI

,

(7)

F ∗ip(s) =∞∑

k=1

L−1∑n=0

n−1∑y=0

L−2∑

j=0

ψ(n, y)pk,(n,y),(L,j) (8)

× λc

s(1− F ∗c (s))(F ∗c (s))k−1

∣∣∣∣s=λI

,

F ∗PA(s) =∞∑

k=1

l−2∑y=0

l−2∑

j=0

ϕ(l−1,y)pk,(l−1,y),(l,j)(F ∗c (s))k

∣∣∣∣s=λI

,

(9)

F ∗pa(s) =∞∑

k=1

l−1∑n=0

n−1∑y=0

l−2∑

j=0

ψ(l, y)pk,(n,y),(l,j) (10)

× λc

s(1− F ∗c (s))(F ∗c (s))k−1

∣∣∣∣s=λI

,

where L and l are the maximum number of the layers for IPpaging area and PA, respectively. Typically, IP paging area is

8

Fig. 7. State transition diagram for Markov chain modeling.

larger than PA, and hence, it satisfies that L ≥ l. We derive theAP-cell crossing probability Pr(Kc = k) simply by replacingF ∗m(s) and F ∗M (s) with λc

s (1− F ∗c (s)) and F ∗c (s), respectively.

D. Power Consumption

We derive MH’s power consumption including WNIC’s whenit adopts the IM or the PSM with IP paging. In [17, 18], the au-thors present the analysis regarding power consumption. We,however, derive a new power consumption model since the pre-viously presented analysis did not deal with hierarchical net-work topology.

Fig. 7 shows two operational states of an MH for a Markovchain modeling by considering our assumptions. The steadystate analysis is based on the Embedded-Markov process be-cause state changes occur with a Markov chain, but take a ran-dom amount of time between the changes. In this figure, p12 andp21 are the state transition probabilities, representing a sessioncompletion in State 1 and a session arrival in State 2, respec-tively. Both p12 and p21 are simply 1, and hence, we easily ob-tain the stationary probabilities of this Embedded Markov Chainas π1 = 1/2 and π2 = 1/2, respectively. In addition, we cananalyze the average time, which the MH stays in each state, asT 1 = 1/λh and T 2 = 1/λI = 1/(λin + λout). Then, we obtainthe steady state probabilities of the Embedded-Markov processas follows:

pi =πiT i∑2

j=1 πjT j

, i = 1, 2. (11)

In the steady state, MH’s average power consumption is de-rived by:

P = p1P 1 + p2P 2,

where p1 and p2 are the probabilities derived in Eq (11) for State1 and State 2, respectively. P 1 and P 2 are the correspondingpower consumption in each state.

Since an MH in the IM does not need to perform IP-layeroperations as discussed in Section III, an HD transits its powermode to standby mode. In contrast, when the PSM with IP pag-ing, is employed, it has to perform inter-AP handoff along withprobable IP-layer handoff. For this reason, it is reasonable toassume HD is always powered on.

We first determine the power P 1 consumed in State 1 as fol-lows:

P 1 = PWN_act + PHD_act,

Table 4. Parameters for Power Consumption Analysis

Parameter Definitionλbr ave. delay for paging message deliveryλin incoming call rateλout outgoing call ratetc a random variable for cell sojourn timeth a random variable for session holding timet(k) a random variable for area sojourn timetr a random variable for residual service timetI a random variable for idle timeDp ave. delay for paging message deliveryE1x energy required to perform 802.1x procedureEAP _res energy required to perform AP reselectionEb energy required to receive beaconEbr energy required to receive broadcastsEDHO energy required for delayed handoffEiho energy required for idle handoffEL2H0 energy required for L2 handoffEL3HO energy required for L3 handoffEp energy required for pagingEW N_slp energy consumed while a WNIC is sleeping when it is in the IMNbi ave. number of beacon listening in IMNbr ave. number of broadcast frames within a Target Beacon Transmis-

sion Time (TBTT)NP A_cng ave. number that an MH leaves PAsP HD_act ave. power consumption of HD being in active stateP HD_slp ave. power consumption of HD being in sleep stateP W N_awk ave. power consumption of WNIC in awake stateP W N_slp min. power consumption of WNIC in doze stateT 1x ave. time for 802.1x procedureT auth ave. time for authentication procedureTbli beacon listening intervalT b ave. time for beacon frame transmissionsT b_sg ave. time for a single beacon frame transmissionT br ave. time for broadcast frame transmissionsT br_sg ave. time for a single broadcast frame transmissionT brp ave. time for inter-arrival of broadcast framesT DHO ave. time for delayed handoff procedureT ih ave. time for idle handoff procedureT L2HO ave. time for L2-level handoffsT L2HO_sg ave. time for a single L2-level handoffT L3HO ave. time for L3-level handoffsT L3HO_sg ave. time for a single L3-level handoffT p ave. time for paging procedureT ras ave. time for reassociation procedureT scan ave. time for scanning procedure

where PWN_act and PHD_act are the power consumptions byactive WNIC and active HD during a session holding time, re-spectively. Prior to further discussion, we summarize all the pa-rameters employed for power consumption analysis in Table 4.

E. Power Consumption for IM

When the IM is employed, P 2 is determined by:

P 2 = P IM + PHD_slp = λIEIM + PHD_slp,

where EIM is the energy consumed for the idle time 1/λI , andis represented by:

EIM =EWN_slp + E1x + Ep + EAP _res

+ Eb +Eiho + EDHO.

Since the 802.1x authentication and delayed handoff proce-dures are performed once at maximum, we approximate EIM

to:

9

EIM 'EWN_slp + Ep + EAP _res + Eb + Eiho

=PWN_slpTWN_slp+

PWN_awk(T pλin

λin + λout+ TAP _res + T b + T iho).

The sleep time TWN_slp becomes the remaining time exclud-ing all the times required for the IM operations. Therefore, wederive TWN_slp by:

TWN_slp

=∫ ∞

0

(t− (T pλin

λin + λout+ T 1x + TL2HO_sg + TAP _res+

T b + T iho + TDHO))× fI(t)dt

=1λI− (T p

λin

λin + λout+ T 1x + TL2HO_sg + TAP _res+

T b + T iho + TDHO),

where T pλin

λin+λoutand T 1x are the times required for pag-

ing and the 802.1x authentication procedures, respectively. L2handoff, composed of scanning, reassociation and authentica-tion, is needed after a successful paging procedure. Accord-ingly, the time TL2HO_sg for a L2 handoff can be obtained bysumming the required times as follows:

TL2HO_sg = T ras + T auth + T scan,

where T ras, T auth, and T scan are the times for reassociation,authentication, and scanning, respectively. TAP _res is the totalrequired time to perform AP reselection during tI . T b, T iho,and TDHO are the times for beacon listening, idle handoff, anddelayed handoff, respectively.

AP reselection is performed at every cell boundary so that wederive TAP _res by obtaining the entire number of cells whichan MH travels during idle time as follows:

TAP _res = T scan

∞∑

k=0

kPr(Kc = k) =T scanF

∗r (λI)

1− F ∗c (λI). (12)

The sum T b of the periodic beacon listening times is deter-mined by:

T b = N biT b_sg,

where N bi is the expected number of beacons which an MHreceives during idle time. Since beacons are received every Tbli,we obtain N bi by dividing idle time by Tbli as follows:

N bi =∫ ∞

0

t

TblifI(t)dt =

1λITbli

.

An MH conducts an idle handoff whenever it crosses a PA sothat the time required for the idle handoffs is derived by:

T iho =T ihF

∗pa(λI)

1− F ∗PA(λI).

Finally, we TWN_slp can be approximated as follows:

TWN_slp ' 1λI− T p

λin

λin + λout

− TL2HO − TAP _res − T b − T iho.

F. Power Consumption for PSM

On the other hand, if the IM is not employed, i.e., PSM withIP paging is used, the power P2 in State 2 is determined by:

P2 = λIEPSM + PHD_act.

We have the energy required for the PSM by:

EPSM =EWN_slp + Eb + Ebr + EL2HO + EL3HO

=PWN_slpTWN_slp (13)

+ PWN_awk(T b + T br + TL2HO + TL3HO),

whereEbr is the energy required to receive broadcast frames. Inorder to support IP paging, MH has to receive as many broad-cast frames as possible since IP paging schemes are designed onthe basis of broadcast frames for the notification of a new callarrival. EL2HO and EL3HO are energies consumed for an L2handoff and an L3 handoff, respectively.

For each term in Eq. (13), we first derive the sleep timeTWN_slp as follows:

TWN_slp =1λI− (T b + T br + TL2HO + TL3HO).

If there exists at least one MH running in the PSM, the serv-ing AP should buffer every broadcast frame and forward it byusing normal transmission rule after DTIM transmission. Forthis reason, under the assumption that broadcast frame arriveswith exponential distribution, we derive the time required to re-ceive the buffered broadcast frames by:

T br = T br_sg

∫ ∞

0

t

TbrpfI(t)dt =

T br_sg

λITbrp,

where Tbrp is expected inter-arrival time of broadcast frames.TL2HO is the total required time for L2 handoffs while an MHcrosses cells. In a similar manner to Eq. (12), we have TL2HO

by:

TL2HO =TL2HO_sgF

∗r (λI)

1− F ∗c (λI).

The time for L3 handoffs TL3HO is derived by:

TL3HO =TL3HO_sgF

∗ip(λI)

1− F ∗IP (λI),

where TL3HO_sg is a time for a single L3 handoff.

10

Table 5. Pr(K = 0) Values1/λc is 160 seconds

1/λI (hour) 1 2 3 4 5 6Pr(K = 0) (numerical) 0.413 0.269 0.200 0.160 0.133 0.114Pr(K = 0) (simulation) 0.414 0.269 0.200 0.161 0.134 0.114

1/λI is 1 hourCell sojourn time (sec) 40 80 160 320 640 1280Pr(K = 0) (numerical) 0.160 0.269 0.413 0.568 0.700 0.811Pr(K = 0) (simulation) 0.161 0.269 0.414 0.566 0.702 0.809

V. ANALYTICAL RESULTS

A. Model Validation

Figs. 8 and 9 show the area crossing probability Pr(K = k),where k ≥ 1, when an MH crosses areas of which the maximumnumber of layers is 6. Table 5 presents the probability that anMH stays at the first area, i.e., k = 0. In order to validate the de-rived equations, we conduct simulation by using the simulator,which we have developed for the exclusive purpose.

Both analysis and simulation results are presented for thecomparison. From both figures and the table, we observe thatanalysis and simulation results match very well, thus verifyingthe validity of our analysis.

Fig. 8 shows Pr(K = k) as the number of area crossingsk increases when 1/λc is fixed at 160 s under the conditionthat E[tI ](= 1/λI) varies from 1 hr to 6 hr. For 0 ≤ k < 4,Pr(K = k) is high when the value of 1/λI is small. In the con-trast, for k > 12, Pr(K = k) is high when 1/λI is long. It im-plies that long 1/λI provides more chances for MH to cross ar-eas as the expectation of idle time 1/λI increases. Fig. 9 showsPr(K = k) under given 1/λI . In this case, the expected cellsojourn time E[tc](= 1/λc) varies from 40 s to 1280 s. Thisfigure shows how mobility influences the area crossing proba-bility. Small 1/λc represents high mobility of an MH. For thisreason, the smaller 1/λc incurs the higher value of Pr(K = k)for k > 10. It implies that smaller 1/λc encourages an MH tocross areas more frequently.

B. Power Consumption Results

Table 6 lists the values of all the parameters used for the nu-merical evaluation including (1) the values from the data sheets(related to the power consumption) [12, 14]8, (2) practicallymeasured values [22, 26]9, and (3) some assumed values. Forsimplicity, we ignore the state transition overhead of WNIC.

Fig. 10 shows the evaluation results about power consumptionand life time according to Average Cell Sojourn Time (ACST).Fig. 10 (a) shows the overall power consumption of WNIC. Land l represent the maximum number of layers of IP paging areaand PA, respectively. For a fair comparison, we use 1 s as well as100 ms for beacon listening interval Tbli. As expected, smallerTbli results in more power consumption.

In this figure, we observe both the IM and PSM with IP pag-ing (PIP) require more power for smaller ACST. As discussedearlier, small ACST implies high mobility. Accordingly, thehigher mobility incurs the more handoffs, thus resulting in the

8PWN_awk is the average of the reception and transmission powers.9In [26], the authors measured the authentication delay when the 802.11i is

employed.

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

2 4 6 8 10 12 14 16 18 20

Pr(

K=

k)

Number of area crossings

Numerical analysis, 1/λI=6hrNumerical analysis, 1/λI=5hrNumerical analysis, 1/λI=4hrNumerical analysis, 1/λI=3hrNumerical analysis, 1/λI=2hrNumerical analysis, 1/λI=1hr

Simulation, 1/λI=6hrSimulation, 1/λI=5hrSimulation, 1/λI=4hrSimulation, 1/λI=3hrSimulation, 1/λI=2hrSimulation, 1/λI=1hr

Fig. 8. Pr(K = k) when 1/λc is 160 seconds.

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

2 4 6 8 10 12 14 16 18 20

Pr(

K=

k)

Number of area crossings

Numerical analysis, 1/λc=40secNumerical analysis, 1/λc=80sec

Numerical analysis, 1/λc=160secNumerical analysis, 1/λc=320secNumerical analysis, 1/λc=640sec

Numerical analysis, 1/λc=1280secSimulation, 1/λc=40secSimulation, 1/λc=80sec

Simulation, 1/λc=160secSimulation, 1/λc=320secSimulation, 1/λc=640sec

Simulation, 1/λc=1280sec

Fig. 9. Pr(K = k) when 1/λI is 1 hour.

Table 6. Parameter Values for Power Consumption Evaluation

Parameter value parameter valueFrom the data sheet

P WN_awk (925+2565)/2 mW P WN_slp 45 mWP HD_act 625 mW P HD_slp 86 mW

Battery cap. 1250 mAh - -From practical measurement

T L3HO 3 s T 1x 525 msAssumed values

T b 500 µs T p 5 msT auth 2 ms Tbli 1 s / 100 msT ras 1.3 ms T ih 500 µsT scan 300 ms 1/λh 5 min

more power consumption. The power consumptions dependingon the mobility are attributed to L2 handoff and AP reselectionin case of PIP and the IM, respectively.

Meanwhile, PIP needs more power in case that L = 4 ratherthan L = 6 under the same ACST. The smaller IP paging areabecomes, the more frequently an MH crosses IP paging areaboundaries, thus consuming more power. However, the power

11

0.04

0.1

0.2

0.5 1 2 4 8 16 32 64

Con

sum

ed p

ower

(W

)

Average cell sojourn time (minutes)

IM, l=6IM, l=4

PIP, tbli=1s, L=6PIP, tbli=1s, L=4

PIP, tbli=100ms, L=6PIP, tbli=100ms, L=4

(a) Power consumption of WNIC

0.5

1

3

6

5 10 20 40 80 160

Life

tim

e (h

ours

)

Average call arrival interval (minutes)

IM, ACST=30secIM, ACST=64min

PIP, tbli=1s, ACST=30secPIP, tbli=1s, ACST=64min

PIP, tbli=100ms, ACST=30secPIP, tbli=100ms, ACST=64min

(b) Life time of MH

Fig. 10. Power consumption and life time.

for the IM remains virtually unchanged irrespective of l. It im-plies the power for idle handoff is ignorable so that the size ofPA hardly influences the amount of the power consumption.

Fig. 10 (b) shows MH’s life time including session holdingtime as well as idle time according to the incoming call arrivalinterval for each ACST bound when l = 4 andL = 6. The upperand the lower bounds of each scheme are obtained for 64 minand 30 sec of ACST, respectively. As expected, higher mobilityresults in shorter life time. However, we observe that the linefor each bound is not well spaced since the power consumptionrequired for an HD as a part of an MH is very large comparedwith that of WNIC.

As a result, we find that the life time become longer as in-coming or outgoing call arrival interval increases. From all theabove observations, we conclude that our scheme provides anMH with longer life time since it needs less power consumptioncompared with the PIP.

VI. CONCLUSION

In this paper, we propose a new protocol to support the IdleMode operation in the 802.11 WLAN. The proposed protocolcan be easily applied to already-deployed products by just up-dating their firmwares or device drivers [15]. It can be employedto work with the existing security schemes including the 802.11rand the 802.11i with minor additional protocol in order to sup-port idle handoff. Therefore, the proposed IM is expected tocontribute to commercial VoIP system deployment by extend-ing the VoIP service life time.

In order to evaluate our proposal, we develop new analyti-cal models, which are useful to analyze the power consumptionin hierarchically structured networks. These network structurewill be typical in near future since wireless access networks areevolving to support IP layer. The numerical results demonstratethat our proposed IM operation outperforms the PSM with IPpaging with respect to the power consumption. As a result, itenables a longer life time of the 802.11-equipped MHs. Addi-tionally, we deal with L3 handoff scheme to reduce the call setup latency incurred by the delayed handoff, and then, integratethem with the proposed protocol for the IM in L2.

APPENDICES

I. PAGING FAILURE PROBABILITY

Let Pf be the paging failure probability. A paging failureoccurs only when a new call arrives before an MH in the IMupdates its location at a new PA. Otherwise, the paging is con-ducted successfully. Therefore, Pf is derived by:

Pf =∞∑

K=1

Pr(K = k)∫ T bli

0

1T bli

∫ T bli

tbli

1T bli

dτdtbli

∫ T bli

0

fc(t)dt

=12

∞∑

K=1

Pr[K = k](1− e−λT bli).

Typically, the average call arrival interval is much longer thanT bli. Therefore, we can justify Pf converges to zero. For ex-ample, when the average call arrival interval is 5 minutes, i.e.,λ = 1/300, and T bli = 1 second, the value of Pf is equal to0.0002.

II. MH’S KTH AREA CROSSING PROBABILITY

Let τ(k) = tm +∑k

i=1 tMi . fτ(k)(t) is the pdf for randomvariable τ(k) and F ∗τ(k)

(s) is its Laplace transform function. Wederive the probability that an MH crosses areas k times duringidle time by:

Pr(K = k) = Pr(τ(k) < tI)− Pr(τ(k+1) < tI)

=∫ ∞

0

∫ ti

0

fI(ti)(fτ(k)(t)− fτ(k+1)(t)

)dtdti

= λI

∫ ∞

0

e−λIti

∫ ti

0

(fτ(k)(t)− fτ(k+1)(t)

)dtdti

= F ∗τ(k)(λI)− F ∗τ(k+1)

(λI)

= F ∗m(λI)(F ∗M (λI))k−1(1− F ∗(λI)),

where k ≥ 1.

12

REFERENCES

[1] IEEE 802.11-1999, Part 11: Wireless LAN Medium Access Control(MAC) and Physical Layer (PHY) specifications, 1999.

[2] IEEE 802.11i, Part 11: Wireless LAN Medium Access Control (MAC) andPhysical Layer (PHY) specifications: Medium Access Control SecurityEnhancements, 2004.

[3] IEEE 802.11k, Part 11: Wireless LAN Medium Access Control (MAC)and Physical Layer (PHY) specifications: Radio Resource Measurement,2008.

[4] IEEE 802.11r, Part 11: Wireless LAN Medium Access Control (MAC)and Physical Layer (PHY) specifications: Fast BSS Transition, 2007.

[5] IEEE 802.11F, Part 11: Wireless LAN Medium Access Control (MAC)and Physical Layer (PHY) specifications: Inter-Access Point Protocol,2003.

[6] S. Jin, C. Yoon, and S. Choi, “A Simple Remedy for Idle Mode via ProxyMIP,” IEEE Communications Letters, June 2008.

[7] M. Shin, A. Mishra, and W. A. Arbaugh, “Improving the Latency of802.11 hand-offs using Neighbor Graphs,” in Proc. ACM MobiSys’04,June 2004.

[8] S. Jin, M. Choi, and S. Choi, “Multiple WNIC-based Handoff in IEEE802.11 WLANs,” IEEE Communications Letters, Oct. 2009.

[9] X. Zhang, J. Gomez, G. Castellanos, and A. T. Campbell, “P-MIP: PagingExtensions for Mobile IP,” ACM Mobile Networks and Applications, July2002.

[10] E. Shih, P. Bahl, and M. J. Sinclair, “Wake on Wireless: An Event DrivenEnergy Saving Strategy for Battery Operated Devices,” in Proc. ACM Mo-biCom’02, Sep. 2002.

[11] T. Ozugur and B. Sarikaya, “Combining Layer 2-Layer 3 Paging for Wire-less LANs,” IEEE Transactions on Wireless Communications, Sep. 2006.

[12] http://www.proxim.com/learn/library/datasheets/11bpccard.pdf.[13] http://www.iana.org/assignments/port-numbers.[14] http://h18000.www1.hp.com/products/quickspecs/11646_na/

11646_na.HTML#TechSpecs(IPAQ h5500).[15] S. Jin, K. Han, and S. Choi, “A Novel Idle Mode Operation in IEEE 802.11

WLANs: Prototype Implementation and Empirical Evaluation,” in Proc.ACM WMASH’06, Sep. 2006.

[16] http://download.intel.com/design/pca/applicationsprocessors/datashts/28000304.pdf.

[17] S. Kwon, S. Nam, H. Hwang, and D. Sung, “Analysis of a Mobility Man-agement Scheme Considering Battery Power Conservation in IP-BasedMobile Networks,” IEEE Transactions on Vehicular Technology, Nov.2004.

[18] Y. Chung, D. Sung, and A. H. Aghvami, “Steady State Analysis of MobileStation State Transition for General Packet Radio Service,” in Proc. IEEEPIMRC’02, Sep. 2002.

[19] Alberto Leon-Garcia, Probability and Random Processes for ElectricalEngineering. New York: Addison-Wesley, 1994, Chapter 9.

[20] Y. Fang, I. Chlamtac, and Y. Lin, “Call Performance for a PCS Network,”IEEE Journ. on Selected Area in Communications, Oct. 1997.

[21] M. Liebsh and X. Pérez-Costa, “Utilization of the IEEE 802.11 PowerSave Mode with IP Paging,” in Proc. IEEE ICC’05, May 2005.

[22] H. Yokota, A. Idoue, T. Hasegawa, and T. Kato, “Link Layer AssistedMobile IP Fast Handoff Method over Wireless LAN Networks,” in Proc.ACM MobiCom’02, Sep. 2002.

[23] I. F. Akyildiz, J. S. M. Ho, and Y. Lin, “Movement-Based Location Up-date and Selective Paging for PCS Networks,” IEEE/ACM Transactions onNetworking, Aug. 1996.

[24] Y. Lin, W. Lai, and R. Chen, “Performance Analysis for Dual Band PCSNetworks,” IEEE/ACM Transactions on Networking, Feb. 2000.

[25] I. F. Akyildiz, Y. Lin, W. Lai, and R. Chen, “A New Random Walk Modelfor PCS Network,” IEEE Journ. on Selected Areas in Communications,July 2000.

[26] S. Bangolae, C. Bell, and E. Qi, “Performance Study of Fast BSS Transi-tion using IEEE 802.11r,” in Proc. ACM IWCMC’06, July 2006.

[27] S. Jin, K. Han, and S. Choi, “A Novel Idle Mode Operation in IEEE 802.11WLANs,” in Proc. ICC’06, June 2006.

[28] A. Mishra, M. Shin, and W. Arbaugh, “Context Caching using NeighborGraphs for Fast Handoffs in a Wireless Network,” in Proc. IEEE INFO-COM’04 Mar. 2004.

Sunggeun Jin is a senior engineer working for ETRI,which he joined in 1998, Korea. Prior to joiningETRI, he received his B.S. and M.S. degrees in Schoolof Electrical Engineering and Computer Science atKyungpook National University (KNU), Korea, in1996 and 1998, respectively. He received his Ph. D. atSchool of Electrical and Computer Engineering, SeoulNational University (SNU), Korea, August, 2008. Hehas participated in standard developments includingIEEE 802.11v, IEEE 802.16j, IEEE 802.16m, andIEEE 802.11ad in serial since 2006. He served as a

TPC member for WCNC 2008, ICUFN 2009, and BROADNETS 2010, and healso completed many peer reviews for journals and conferences such as IEEETMC, IEEE INFOCOM, IEEE ICC, IEEE GLOBECOM, and IEEE WCNC. Heis now studying directional MAC and power saving strategies in 60 GHz band.

Kwanghun Han received his B.E. degree in theSchool of Electronic Engineering from Seoul NationalUniversity (SNU), Seoul, Korea in February 2004. Heis currently working toward his Ph.D. degree at Schoolof Electrical Engineering, SNU. His research inter-ests include radio resource allocation and optimiza-tion, power saving in wireless networks, and MAC de-sign for emerging systems.

Sunghyun Choi is currently an visiting associate pro-fessor at Stanford University, USA, and an associateprofessor at the School of Electrical Engineering,Seoul National University (SNU), Seoul, Korea. Be-fore joining SNU in September 2002, he was withPhilips Research USA, Briarcliff Manor, New York,USA as a Senior Member Research Staff and a projectleader for three years. He received his B.S. (summacum laude) and M.S. degrees in electrical engineeringfrom Korea Advanced Institute of Science and Tech-nology (KAIST) in 1992 and 1994, respectively, and

received Ph.D. at the Department of Electrical Engineering and Computer Sci-ence, The University of Michigan, Ann Arbor in September, 1999. His cur-rent research interests are in the area of wireless/mobile networks with empha-sis on wireless LAN/MAN/PAN, next-generation mobile networks, mesh net-works, cognitive radios, resource management, data link layer protocols, andcross-layer approaches. He authored/coauthored over 120 technical papers andbook chapters in the areas of wireless/mobile networks and communications.He has co-authored (with B. G. Lee) a book "Broadband Wireless Access andLocal Networks: Mobile WiMAX and WiFi," Artech House, 2008. He holds15 US patents, nine European patents, and nine Korea patents, and has tensof patents pending. He has served as a General Co-Chair of COMSWARE2008, and a Technical Program Committee Co-Chair of ACM Multimedia 2007,IEEE WoWMoM 2007 and IEEE/Create-Net COMSWARE 2007. He was aCo-Chair of Cross-Layer Designs and Protocols Symposium in IWCMC 2006,2007, and 2008, the workshop co-chair of WILLOPAN 2006, the General Chairof ACM WMASH 2005, and a Technical Program Co-Chair for ACM WMASH2004. He has also served on program and organization committees of numerousleading wireless and networking conferences including IEEE INFOCOM, IEEESECON, IEEE MASS, and IEEE WoWMoM. He is also serving on the edito-rial boards of IEEE Transactions on Mobile Computing, ACM SIGMOBILEMobile Computing and Communications Review (MC2R), and Journal of Com-munications and Networks (JCN). He is serving and has served as a guest editorfor IEEE Journal on Selected Areas in Communications (JSAC), IEEE Wire-less Communications, Pervasive and Mobile Computing (PMC), ACM WirelessNetworks (WINET), Wireless Personal Communications (WPC), and WirelessCommunications and Mobile Computing (WCMC). He gave a tutorial on IEEE802.11 in ACM MobiCom 2004 and IEEE ICC 2005. From 2000 to 2007, hewas a voting member of IEEE 802.11 WLAN Working Group. He has receiveda number of awards including the Young Scientist Award awarded by the Pres-ident of Korea (2008); IEEK/IEEE Joint Award for Young IT Engineer (2007);the Outstanding Research Award (2008) and the Best Teaching Award (2006)both from the College of Engineering, Seoul National University; the Best Pa-per Award from IEEE WoWMoM 2008; and Recognition of Service Award(2005, 2007) from ACM. Dr. Choi was a recipient of the Korea Foundationfor Advanced Studies (KFAS) Scholarship and the Korean Government Over-seas Scholarship during 1997-1999 and 1994-1997, respectively. He is a seniormember of IEEE, and a member of ACM, KICS, IEEK, KIISE.