IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. X, NO. X .

2y ago
100 Views
2 Downloads
1.20 MB
14 Pages
Last View : 12d ago
Last Download : 2m ago
Upload by : Matteo Vollmer
Transcription

IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. X, NO. X, XXXXX 20XX1On the Energy Efficiency of Device Discovery inMobile Opportunistic Networks: A SystematicApproachBo Han, Jian Li and Aravind Srinivasan Fellow, IEEEAbstract—In this paper, we propose an energy efficient device discovery protocol, eDiscovery, as the first step to bootstrappingopportunistic communications for smartphones, the most popular mobile devices. We chose Bluetooth over WiFi as the underlyingwireless technology of device discovery, based on our measurement study of their operational power at different states on smartphones.eDiscovery adaptively changes the duration and interval of Bluetooth inquiry in dynamic environments, by leveraging historyinformation of discovered peers. We implement a prototype of eDiscovery on Nokia N900 smartphones and evaluate its performancein three different environments. To the best of our knowledge, we are the first to conduct extensive performance evaluation of Bluetoothdevice discovery in the wild. Our experimental results demonstrate that compared with a scheme with constant inquiry duration andinterval, eDiscovery can save around 44% energy at the expense of discovering only about 21% less peers. The results also showthat eDiscovery performs better than other existing schemes, by discovering more peers and consuming less energy. We also verifythe experimental results through extensive simulation studies in the ns-2 simulator.Index Terms—Device discovery, opportunistic communications, energy efficiency, smartphones, Bluetooth. 1I NTRODUCTIONMobility itself is a significant problem in mobile networking. On the one hand, protocols designed for mobilenetworks should solve the challenges caused by the mobility of wireless devices. For example, routing protocols,such as DSR (Dynamic Source Routing) [16], are requiredto handle frequent routing changes and reduce the corresponding communication overhead. On the other hand,mobility can increase the capacity of wireless networksthrough opportunistic communications [13], where mobile devices moving into wireless range of each othercan exchange information opportunistically during theirperiods of contact [7], [21].Opportunistic communications have been widely explored in delay-tolerant networks [32], mobile socialapplications [21], [31] and mobile advertising [1], to facilitate message forwarding, media sharing and locationbased services. Meanwhile, there are more and more applications leveraging opportunistic communications forvarious purposes. For example, LoKast1 is an iPhone application that provides mobile social networking services B. Han is with AT&T Labs – Research, 1 AT&T Way, Bedminster, NJ,07921. This work was done when he was a graduate student at theUniversity of Maryland.E-mail: bohan@research.att.com J. Li is with the Institute for Interdisciplinary Information Sciences,Tsinghua University, Beijing 100084, China.E-mail: lijian83@mail.tsinghua.edu.cn A. Srinivasan is with the Department of Computer Science and theInstitute for Advanced Computer Studies, University of Maryland, CollegePark, MD, 20742.E-mail: srin@cs.umd.edu1. http://www.lokast.com/by discovering and sharing media content among usersin proximity. Nintendo 3DS’s StreetPass2 enables playersto exchange game data with other users they pass onthe street, through the direct device-to-device communication between 3DS systems. Other similar applicationsinclude Sony PS Vita’s Near and Apple’s iGroups.Device discovery is essentially the first step of opportunistic communications. However, there are veryfew practical protocols proposed for it and most ofthe existing work mainly utilizes (trace-driven) simulation to evaluate the performance of various devicediscovery protocols [9], [29]. Moreover, although thereare several real-world mobility traces in the CRAWDADrepository3 which were collected using Bluetooth devicediscovery, most of them used very simple discoveryprotocols with constant inquiry duration and interval. Arecently proposed opportunistic Twitter application [24]also uses a 2-minute inquiry interval for Bluetooth devicediscovery. It is known that these kinds of discoveryprotocols are not energy efficient [29] and thus maynot be desirable for power-constrained mobile devices,such as smartphones. In this paper, we bridge this gapby developing an energy-aware device discovery protocol for smartphone-based opportunistic communicationsand evaluating its performance in practice.There are two major challenges in designing, implementing and evaluating energy efficient device discoveryprotocols for smartphones. First, the selection of underlying communication technology is complicated bythe multiple wireless interfaces on smartphones, such2. http://www.nintendo.com/3ds/features/3. http://crawdad.cs.dartmouth.edu/

2IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. X, NO. X, XXXXX 20XXas Bluetooth and WiFi (a.k.a., IEEE 802.11).4 AlthoughBluetooth is a low-power radio, its device discoveryduration is much longer than WiFi ( 10s for Bluetoothvs. 1s for WiFi active scanning), which may cause moreenergy consumption on smartphones. Similarly, WiFi isknown to be power-hungry for mobile devices [22], [26].Thus, it is not clear which of them is more suitable fordevice discovery on smartphones.Second, given the dynamic nature of human mobility,we need to adaptively tune device discovery parameters, such as inquiry duration and interval, to reducesmartphone energy consumption. Schemes with constantinquiry intervals have been proven to be optimal forminimizing discovery-missing probability [29]. However, their energy consumption is usually higher than theadaptive ones, which may miss more devices during discovery procedures. Therefore, there is a tradeoff betweenenergy consumption and discovery-missing probability.We make the following contributions in this paper. We present a systematic measurement study of theenergy consumption of Bluetooth and WiFi devicediscovery on smartphones, by measuring both theelectrical power at various states and the discovery duration (Section 4). Our results show thatthe energy consumption depends on the numberof discovered peers. Based on our measurementresults, we chose Bluetooth as the underlying wireless technology, because even its high-power stateconsumes less energy than the low-power state ofWiFi during device discovery. We emphasize thatalthough previous works have studied the powerof Bluetooth/WiFi devices [9], [11], [22], they eitherfocus on only Bluetooth [9] or ignore the duration ofdevice discovery [11], [22], without which it is hardto evaluate the energy consumption of these devices. We design an energy-aware device discovery protocol, named eDiscovery, as the first and very important step to bootstrapping smartphone-based opportunistic communications (Section 6). By tradingenergy consumption for a limited discovery loss, wedemonstrate that eDiscovery is highly effectivein saving energy on smartphones. eDiscovery dynamically tunes the discovery duration and intervalaccording to history information of the number ofdiscovered peers. It also introduces randomizationinto device discovery, in order to explore the searchspace further. Our major contribution is an extensive performanceevaluation of eDiscovery and other existing device discovery protocols in different realistic environments, through a prototype implementation onNokia N900 smartphones (Section 7). We conductexperiments in a university campus, a metro stationand a shopping center. Our experimental resultsverify the effectiveness of eDiscovery in practice.Compared with the STAR protocol proposed byWang et al. [29], eDiscovery consumes less energyand discovers more peers. eDiscovery also performs better than another protocol in the literature.Compared with its preliminary version [15], this paper makes two new contributions. First, we add thetheoretical analysis of device-discovery missing probability (Section 5), which motivates us to design theeDiscovery protocol. Second, we port the implementation of eDiscovery into the ns-2 simulator enhancedwith the UCBT Bluetooth module5 and offer a detailedevaluation of its parameters (Section 8). We also comparethe performance of eDiscovery and STAR with morenetwork topologies using ns-2 simulations.4. We prefer Bluetooth and WiFi to 3G, as they are localcommunication technologies with almost no monetary cost.2 D EVICE D ISCOVERY IN B LUETOOTH ANDWIFIIn the following, we discuss device discovery of Bluetooth and WiFi, the two most commonly available localwireless communication technologies on smartphones.2.1 BluetoothThe Bluetooth specification (Version 2.1) [3] defines alllayers of a typical network protocol stack, from thebaseband radio layer to the application layer. Bluetoothoperates in the 2.4 GHz ISM (Industrial, Scientific andMedical) frequency band, shared with other devices suchas IEEE 802.11 stations, baby monitors and microwaveovens [12]. Therefore, it uses Frequency-Hopping SpreadSpectrum (FHSS) to avoid cross-technology interference,by randomly changing its operating frequency bands.Bluetooth has 79 frequency bands (1 MHz width) in therange 2402-2480 MHz and the duration of a Bluetoothtime slot is 625 µs. In the following we focus on devicediscovery and refer interested readers to Smith et al. [27]for further study of the Bluetooth protocol stack.During device discovery, an inquiring device sendsout inquiry messages periodically and waits for responses, and a scanning device listens to wirelesschannels and sends back responses after receiving inquiries [3]. The inquiring device uses two trains of 16frequency bands each, selected from 79 bands. The 32bands of these two trains are selected according to apseudo-random scheme and a Bluetooth device switchesits trains every 2.56 seconds. In every time slot, theinquiring device sends out two inquiry messages ontwo different frequency bands and waits for responsemessages on the same frequency bands during the nexttime slot. After a device receives an inquiry message, itwill wait for 625 µs (i.e., the duration of a time slot)before sending out a response message on the samefrequency band, which completes the device discoveryprocedure. For scanning devices, Bluetooth controls theirscanning duration and frequency with two parameters,scan window and scan interval.5. http://www.cs.uc.edu/ cdmc/ucbt/

HAN et al.: ON THE ENERGY EFFICIENCY OF DEVICE DISCOVERY IN MOBILE OPPORTUNISTIC NETWORKS: A SYSTEMATIC APPROACH2.2Bluetooth Low EnergyBluetooth Low Energy (LE) [4] operates in the 24002483.5 MHz frequency band and divides this band into40 channels with 2 MHz width, instead of 79 channelswith 1 MHz width in the classic Bluetooth. Three out ofthese 40 channels, with channel indexes 37, 38, and 39,are used for advertising, and the rest are data channels.Differently from the classic Bluetooth, the LE systemleverages these advertising channels for device discoveryand connection establishment. Among the five statesdefined in Bluetooth LE, three of them are related todevice discovery: advertising, scanning and initiatingstates (the rest two are standby and connection states).After a device enters the advertising state (directed bythe host machine), it sends out one or more advertising packets that contains its device address on theadvertising channels. These advertising packets composethe so-called advertising events. The time between thestart of two consecutive advertising events is defined asthe sum of a constant advInterval, which should be aninteger multiple of 625 µs and in the range of 20 msto 10.24 s, and a pseudo-random value advDelay in therange of 0 ms to 10 ms. A device in either scanning orinitiating state listens on an advertising channel with theduration of scanWindow and the interval scanInterval (i.e.,the interval between the start of two consecutive scanwindows). The scanWindow and scanInterval parametersshould not be greater than 10.24 s.2.3WiFiThe key concept of device discovery in WiFi is wellunderstood. WiFi stations in infrastructure and ad-hocmodes periodically (100 ms by default) send out Beacon messages to announce the presence of a network.A Beacon message includes information such as SSID(service set identifier) and capability information. TheWiFi interfaces of mobile phones should operate in adhoc mode and form an Independent Basic Service Set(IBSS) to support opportunistic communications, sinceinfrastructure-mode interfaces cannot form a networkand thus cannot communicate directly. Besides sendingout Beacon messages, a WiFi interface also scans wirelesschannels to discover peers.There are two types of WiFi scanning, passive and active. In passive scanning, a WiFi interface listens for Beacon messages on each channel, broadcasted by its peersat regular intervals. It periodically switches channels,but does not send any probe request message. Duringactive scanning, a WiFi interface actively searches for itspeers, by broadcasting probe request messages on eachpossible operating channel (channels 1 to 11 in NorthAmerica). It then waits for probe response messagesfrom its peers, which include information similar to thatin Beacon messages.We prefer active scanning to passive scanning fordevice discovery of opportunistic networks for two reasons. First, although passive scanning has the advantage3of not broadcasting probe request messages, it dwellson each channel longer than active scanning, to collectBeacon messages from peers, and thus may consumemore energy. Second, an ad-hoc mode interface may skipthe sending of Beacon messages and thus make itselfnot discoverable by passive scanning, when it tries toscan for other peers with the same SSID (which happensfrequently when it is the only station in an IBSS).3R ELATED W ORKIn this section, we briefly review the literature of devicediscovery in wireless networks.3.1 Wireless Device Discovery in GeneralDevice discovery has been studied in various wirelessnetworks, such as ad-hoc networks [20], [28], sensornetworks [10], [17] and delay-tolerant networks [29].Neighbor/device discovery is one of the first stepsto initialize large wireless networks. McGlynn and Borbash [20] examine the problem of neighbor discoveryduring the deployment of static ad-hoc networks, wherethe discovery may last only a few minutes. Vasudevanet al. [28] show that an existing ALOHA-like neighbordiscovery algorithm reduces to the classical Coupon Collector’s Problem when nodes are not capable of collisiondetection. They also propose an improved algorithmbased on receiver status feedback when nodes have acollision detection mechanism. Cohen and Kapchits [6]investigate a slightly different neighbor discovery problem in asynchronous sensor networks. Instead of studythe initial neighbor discovery, they are interested incontinuous neighbor discovery after the initial discoveryphase. Unlike the above works that are based on abstractcommunication models, our focus is practical Bluetoothdevice discovery for smartphone-based opportunisticcommunications.Dutta and Culler [10] propose an asynchronous neighbor discovery protocol, called Disco, for mobile sensingapplications. U-Connect [17] is another asynchronousneighbor discovery protocol for mobile sensor networksthat selects carefully the time slots to perform discoveryand that has been proven theoretically better than Disco.Recently, Bakht et al. [2] propose Searchlight, a protocol that combines both deterministic and probabilistic approaches to further reduce the discovery latencyfor mobile social applications. Disco, U-Connect andSearchlight mainly aim to achieve a tradeoff betweendiscovery latency and energy consumption. For example,U-Connect uses the power-latency product metric forperformance evaluation. Differently from them, we areinterested in the tradeoff between energy consumptionand discovery-missing probability.The goal of eDiscovery is similar in spirit to thatof Wang et al. [29] who investigate the tradeoff between the contact probing frequency (which determinesenergy consumption) and the missing probability of acontact for delay tolerant applications. They design a

4IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. X, NO. X, XXXXX 20XXcontact probing algorithm, named STAR (Short TermArrival Rate), to dynamically change the contact probingfrequency. Specifically, STAR estimates the peer arriving rate over a short time, which decays slowly, andcalculates the probing frequency based on the estimation. Without specifying the communication technologies, they assume that every probing message is just animpulse and consumes no time. We compare the performance of eDiscovery with STAR through extensivereal-world experiments and simulation studies.FlashLinQ [30] is a synchronous wireless networkarchitecture developed by Qualcomm for direct deviceto-device communication over licensed spectrum. Although FlashLinQ may be more energy efficient thanBluetooth and WiFi, given its clean-slate design for adhoc networks, it requires special purpose hardware andalso operates in licensed spectrum. Unlike from FlashLinQ, we aim to design and implement device discoveryprotocols using existing hardware and communicationtechnologies available on commercial smartphones.an energy-aware device discovery protocol in the wildthrough a prototype implementation on smartphones.Bluetooth device discovery has been an importantcomponent in opportunistic networks. We refer interested readers to a preliminary version of this paper [15]for a literature review in that area.3.2Bluetooth Device DiscoveryBluetooth specifies a detailed device discovery protocol [3]. Salonidis et al. [25] identify the bottlenecksof asymmetric device-discovery delay of Bluetooth andintroduce a randomized symmetric discovery protocol toreduce this delay. Based on Bluetooth specification v1.1,Peterson et al. [23] derive rigorous expressions for theinquiry-time probability distribution of two Bluetoothdevices that want to discover each other. Chakrabortyet al. [5] present an analytical model of the time ofBluetooth device discovery protocol. Liberatore et al. [18]solve the problem of long discovery duration of Bluetooth due to its half-duplex discovery process by theaddition of another Bluetooth radio.Drula et al. [9] study how to select Bluetooth devicediscovery parameters according to the mobility contextand thus reduce the energy consumption of device discovery. They present two algorithms that adjust theseparameters based on recent activity level (referred asRAL) and the location of previous contacts, and evaluatetheir performance through simulations. RAL defines several inquiry modes based on parameters, such as inquiryduration and interval, and switches to a more aggressivemode whenever another device is discovered. In ourprevious work [14], we compare energy consumptionof Bluetooth and WiFi device discovery on Nokia N900smartphones, using battery life as a metric. We evaluatethe Bluetooth device-discovery probability in an officeenvironment using a static phone and a moving phone.Besides the above works, although there is a largebody of literature about Bluetooth device discovery, mostof them focus on the improvements of discovery latency between two Bluetooth devices by tuning variousparameters or changing the protocol itself, which maynot be feasible to implement on smartphones. The focusof this paper is about the performance evaluation of4E NERGY C ONSUMPTION OF D EVICE D IS -COVERYIn this section, we measure the power and energyconsumption of Bluetooth and WiFi device discoveryon smartphones. Based on the experimental results, wechose Bluetooth as the communication technology. Although previous work has measured energy consumption of WiFi and Bluetooth devices several years ago [9],[22], these results may be invalid given the rapid development of battery and wireless technologies [11]. Friedman et al. [11] have recently studied the power of Bluetooth scanning and WiFi search. However, they overlook the duration of device discovery which determinesthe energy consumption on smartphones. Furthermore,their measurements are for station mode WiFi interfacesand demonstrate inconsistent results about WiFi devicediscovery. To the best of our knowledge, there is nosystematic study of smartphone energy consumption ofBluetooth and WiFi device discovery.4.1 Measurement SetupWe measure the electrical power of two states of Bluetooth and WiFi device discovery, idle (i.e., inquiry interval) and active probing, on Nokia N900 smartphonesusing the Monsoon power monitor6 . The default OS ofNokia N900, Maemo 5, is an open source Linux distribution (kernel version 2.6.28). Its WiFi chipset is TexasInstruments WL1251 using the wl12xx device driver7 . ItsBluetooth chipset is Broadcom BCM2048. We use BlueZ8 ,the default Bluetooth protocol stack of most Linux distributions, to run Bluetooth device discovery experiments.During the measurements, we redirect standard outputto \dev\null and turn the screen off to minimizetheir impact on the measurement results. We report theaverage result and the 95% confidence intervals for eachconfiguration over 10 runs in this section. Note that allresults of power measurements in this paper include thebaseline power of the smartphone under test.4.2 BluetoothWe present a 60-second snapshot of the power of Bluetooth device discovery in Figure 1. We perform theexperiments by running hcitool, a tool that can sendcommands, such as inq (inquiry), to Bluetooth devices.We use the flush option to clear the cache of previously6. http://www.msoon.com/LabEquipment/PowerMonitor/7. http://linuxwireless.org/en/users/Drivers/wl12xx8. http://www.bluez.org/

180018001500150012001200Power (mW)Power (mW)HAN et al.: ON THE ENERGY EFFICIENCY OF DEVICE DISCOVERY IN MOBILE OPPORTUNISTIC NETWORKS: A SYSTEMATIC APPROACH90060030090060030000010203040506001020Time (s)Fig. 1: A 60-second snapshot of the temporal power of periodic Bluetooth device discovery with 10-second interval.The smartphone under test is a Nokia N900 smartphone.# of DevicesAverage95% Confidence Interval0162.03(160.72, 163.34)1227.06(219.42, 234.70)2247.72(242.39, 253.05)3248.91(243.02, 254.80)4248.59(246.63, 250.55)5256.02(252.96, 259.08)6253.05(249.63, 256.47)TABLE 1: The electrical power (in mW) of Bluetoothdevice discovery with different numbers of peers.discovered devices before each inquiry. During the measurements, the phone queries neighboring Bluetooth devices periodically with a 10-second interval. When thereis no neighboring device, the average power of Bluetoothinquiry over 10 runs is 162.03 mW (standard deviation:2.12 mW). During the idle states, the Bluetooth radio isin discoverable mode with average power 16.54 mW(standard deviation: 1.11 mW).The average power of Bluetooth device discoveryis affected by the number of neighboring devices. Werepeat the experiments with the number of neighboringBluetooth devices increasing from 0 to 6 and summarizethe results in Table 1. As we can see from this table,when there is one neighboring device, the average powerincreases to around 227.06 mW, due to the reception ofresponse messages of Bluetooth inquiry. When there aremore than one neighboring devices, the average powerincreases to about 250 mW. Defined in the standard [3],the duration of Bluetooth device discovery should be amultiple of 1.28 seconds and the recommended defaultvalue is 10.24 seconds, which we used in the measurements. Figure 1 shows clearly the configured Bluetoothdevice discovery duration and interval.4.35WiFiWe present another 60-second snapshot of the powerof WiFi device discovery in Figure 2. We perform the30405060Time (s)Fig. 2: A 60-second snapshot of the temporal power ofperiodic WiFi device discovery with 10-second interval.The smartphone under test is a Nokia N900 smartphone.EnvironmentOfficeHomePark# of peers43.52 (42.48, 44.56)14.02 (13.75, 14.29)0.01 (n/a)duration (s)1.07 (1.04, 1.10)0.87 (0.86, 0.88)0.52 (0.51, 0.53)TABLE 2: The average number of discovered peers andduration of WiFi device discovery in three environments.The numbers in the parentheses are the 95% confidenceintervals.experiments by running iwlist, a tool that shows thelist of access points and ad-hoc cells in range throughactive scanning. During the measurements, the phonescans neighboring devices periodically also with a 10second interval, which can be clearly identified in Figure 2. The average power of WiFi active scanning over 10runs is 836.65 mW (standard deviation: 8.98 mW). Evenduring scanning intervals, the average power is 791.02mW (standard deviation: 5.23 mW), because the WiFiradio is in ad-hoc mode and sends out Beacon messageswith 100 ms intervals.Differently from Bluetooth, the duration of WiFi activescanning is not constant and may depend on the numberof operation channels and the amount of neighboringpeers. We measure the duration of WiFi device discoveryin three different environments: a campus office building, an apartment, and a national park, and summarizethe results in Table 2. In each environment, we repeat theexperiments 100 times and report the average values andthe 95% confidence intervals. As we can see from thistable, when the number of discovered peers increases,the duration of WiFi device discovery grows from 0.52seconds to 1.07 seconds, which is much shorter thanthe duration of Bluetooth inquiry.4.4 Energy ConsumptionWe summarize the average power of Bluetooth (with6 neighboring devices) and WiFi device discovery inTable 3. Suppose the power is Pidle for the idle stateand Pprobe for the inquiry/scan state of Bluetooth/WiFidevices, the duration of Bluetooth inquiry/WiFi scan is

6180018001500150012001200Power (mW)Power (mW)IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. X, NO. X, XXXXX 20XX90060030090060030000010203040506001020Time (s)Fig. 3: A 60-second snapshot of the temporal power of periodic Bluetooth device discovery with 10-second interval.The smartphone under test is a HTC Hero smartphone(Android 1.5).BluetoothWiFiPidle16.54 (15.85, 17.23)791.02 (787.78, 794.26)30405060Time (s)Pprobe253.05 (249.63, 256.47)836.65 (831.08, 842.22)TABLE 3: The average power of Bluetooth and WiFidevice discovery in mW. The numbers in the parenthesesare the 95% confidence intervals.Tprobe and the inquiry/scan interval is Tidle . Then theestimated energy consumption isE Tidle · Pidle Tprobe · PprobeGiven the high power of WiFi device discovery in bothactive probing and idle states, we prefer Bluetooth toWiFi for device discovery of smartphone-based opportunistic communications. We note that no matter howlong the duration of Bluetooth inquiry is, the overall energy consumption of Bluetooth device discovery shouldalways be lower than that of WiFi, because the power ofBluetooth inquiry is even lower than that of the WiFi idlestate (253.05 vs. 791.02 mW). To perform device discovery, the major problem of WiFi ad-hoc mode is that theradio needs to send out Beacon messages periodicallyand power saving mechanisms for WiFi ad-hoc modeare not available on most mobile phones [26].Although the communication range of WiFi is longerthan Bluetooth and may discover more peers, makingits device discovery energy efficient requires substantialmodifications of the WiFi protocol, which may not befeasible on most smartphones. In this paper, we aimto design a device discovery protocol without changingthe underlying communication protocol and thus makeits deployment easy. This is another reason why wechose Bluetooth over WiFi. However, we emphasize thatif energy consumption is not a major concern and thedesign goal is to discover as many peers as possible or totransfer a large amount of data efficiently, we should useWiFi as the underlying communication protocol (whichis out of the scope of this paper), because it has a largercoverage area.Fig. 4: A 60-second snapshot of the temporal power ofperiodic WiFi device discovery with 10-second interval.The smartphone under test is a HTC Hero smartphone(Android 1.5).4.5 Android SmartphonesWe also measured the power of Bluetooth and WiFidevice discovery using a HTC Hero smartphone withAndroid 1.5. We plot the results in Figure 3 for Bluetoothand Figure 4 for WiFi. On this smartphone, the averagepower is 432.84 mW (standard deviation: 7.86 mW) forBluetooth inquiry and 900.25 mW (standard deviation:21.54 mW) for WiFi scan. There are two differences ofthe experiments on the Nokia N900 and HTC Herosmartphones. First, the experiments on HTC Hero wereperformed with the screen on due to the operationalrequirements and thus the baseline power of HTC Herois higher than that of Nokia N900. Second, the WiFiinterface on HTC Hero does not support ad-hoc modeand we cannot measure the average power Pidle on it.However, these results still clearly show the significantpower difference (467.41 mW) of Bluetooth inquiry andWiFi scan.5D EVICE D ISCOVERY M ISSING P ROBABILITYIn this section, we analyze the missing probability ofa device discovery protocol with constant Bluetoothinquiry window and interval (referred as Constant inthe following).First, we introduce some notations. For a given devicei, we assume that the contact durations tD (i) are independent and identically distributed random variableswith common PDF (probability density function) p(x) ddx Pr[tD x]. We assume the inter-contact time (thetime between subsequent contacts) tC (i) are stationaryrandom variables.If a scanning device is in the discovery/contact rangeof an inquiring device for L seconds, it can be discoveredwith probability R(L). We can easily derive R(L) fromthe analysis of the probability distribution of the inquirytime for Bluetooth devices by Peterson et al. [23]. ForBluetooth device discovery, R(L) is a monotonicallyincreasing function of L. Let R(L) 1 R(L). Assumethat for different devices, or for the same device in

HAN et al.: ON THE ENERGY EFFICIENCY OF DEVICE DISCOVERY IN MOBILE OPPORTUNISTIC NETWORKS: A SYSTEMATIC

Nokia N900 smartphones (Section 7). We conduct experiments in a university campus, a metro station and a shopping center. Our experimental results verify the effectiveness of eDiscoveryin practice. 4. We prefer Bluetooth and WiFi to 3G, as they are local communication technologies with alm

Related Documents:

IEEE 3 Park Avenue New York, NY 10016-5997 USA 28 December 2012 IEEE Power and Energy Society IEEE Std 81 -2012 (Revision of IEEE Std 81-1983) Authorized licensed use limited to: Australian National University. Downloaded on July 27,2018 at 14:57:43 UTC from IEEE Xplore. Restrictions apply.File Size: 2MBPage Count: 86Explore furtherIEEE 81-2012 - IEEE Guide for Measuring Earth Resistivity .standards.ieee.org81-2012 - IEEE Guide for Measuring Earth Resistivity .ieeexplore.ieee.orgAn Overview Of The IEEE Standard 81 Fall-Of-Potential .www.agiusa.com(PDF) IEEE Std 80-2000 IEEE Guide for Safety in AC .www.academia.eduTesting and Evaluation of Grounding . - IEEE Web Hostingwww.ewh.ieee.orgRecommended to you b

Signal Processing, IEEE Transactions on IEEE Trans. Signal Process. IEEE Trans. Acoust., Speech, Signal Process.*(1975-1990) IEEE Trans. Audio Electroacoust.* (until 1974) Smart Grid, IEEE Transactions on IEEE Trans. Smart Grid Software Engineering, IEEE Transactions on IEEE Trans. Softw. Eng.

IEEE TRANSACTIONS ON CLOUD COMPUTING, VOL. XX, NO. XX, MONTH YEAR 1 Cost-Aware Multimedia Data Allocation for Heterogeneous Memory Using Genetic Algorithm in Cloud Computing Keke Gai, Student Member, IEEE, Meikang Qiu, Member, IEEE, Hui Zhao Student Member, IEEE Abstract—Recent expansions of Internet-of-Things (IoT) applying cloud computing .

IEEE TRANSACTIONS ON IMAGE PROCESSING, TO APPEAR 1 Quality-Aware Images Zhou Wang, Member, IEEE, Guixing Wu, Student Member, IEEE, Hamid R. Sheikh, Member, IEEE, Eero P. Simoncelli, Senior Member, IEEE, En-Hui Yang, Senior Member, IEEE, and Alan C. Bovik, Fellow, IEEE Abstract— We propose the concept of quality-aware image, in which certain extracted features of the original (high-

IEEE Robotics and Automation Society IEEE Signal Processing Society IEEE Society on Social Implications of Technology IEEE Solid-State Circuits Society IEEE Systems, Man, and Cybernetics Society . IEEE Communications Standards Magazine IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology IEEE Transactions on Emerging .

EIC, IEEE Transactions on Cloud Computing – Yuanyuan Yang EIC, IEEE Transactions on Cognitive Communications and Networking – Ying-Chang Liang EIC, IEEE Transactions on Molecular, Biological, and Multi-Scale Communications – Chan-Byoung Chae EIC, IEEE Transactions on Signal and Info

Standards IEEE 802.1D-2004 for Spanning Tree Protocol IEEE 802.1p for Class of Service IEEE 802.1Q for VLAN Tagging IEEE 802.1s for Multiple Spanning Tree Protocol IEEE 802.1w for Rapid Spanning Tree Protocol IEEE 802.1X for authentication IEEE 802.3 for 10BaseT IEEE 802.3ab for 1000BaseT(X) IEEE 802.3ad for Port Trunk with LACP IEEE 802.3u for .

IEEE Transactions on Cloud Computing IEEE TRANSACTIONS ON CLOUD COMPUTING 1 xxxx Application-Aware Big Data Deduplication in Cloud Environment . Yinjin Fu, Nong Xiao, Hong Jiang, Fellow, IEEE, Guyu Hu, and Weiwei Chen. Abstract —Deduplication has become a widely deployed technology in cloud