PERVASIVE APPLICATION PROFILING

Mobile Application Profiling for Connected Mobile Devices

The SymPA ( Protocol Analyzer) tool correlates traffic information, radio-access-technology measurements, and location data to help developers evaluate mobile applications in the field.

he recent appearance of open has different behaviors����������������������������������������������������������������� and is suitable for differ- operating systems and virtual ent contexts. In addition, the simultaneous use machines optimized for mobile of different radio-access technologies is a new devices has led to new and inno- challenge for mobile developers. vative applications.1 However, Several existing tools can analyze a mobile ap- Talthough these mobile devices pave the way for a plication’s resource consumption performance. new generation of applications that exploit their However, these tools don’t address communica- intrinsic mobility, proximity to the user, and tion performance from a developer’s standpoint dependence on handheld devices, they create (see the “Related Work in Application Profil- unique challenges for developers.2 For instance, ing” sidebar) or answer these questions: new mobile devices come with powerful multi- tasking processors, but their performance and • How can I debug the application’s data con- processing capacity differ significantly from nection behavior? those of traditional desktop • What’s my protocol’s performance? Almudena Díaz, Pedro Merino, computers. In this restricted • Can I improve my application’s connection and F. Javier Rivas and unknown development performance? University of Málaga environment, developers face • What’s my application’s connection quality? new constraints related to not • What’s my real available bandwidth? only hardware issues such as processors and memory but also the lack of Such questions proliferate in dynamic scenarios tools for debugging the different aspects that in which propagation conditions are less stable affect mobile application performance. and vertical handover occurs between different Mobile device technologies such as , radio-access technologies. wireless local area networks (WLANs), General To provide the answers to these and other Packet Radio Service (GPRS) or Enhanced Data questions, and to help developers find the Rate for Global Evolution (EDGE), Universal cause of errors related to accessing network Mobile Telecommunications System (UMTS)/ resources, we created SymPA (Symbian Proto- High-Speed Packet Data Access (HSDPA), and col Analyzer; www.lcc.uma.es/~pedro/mobile). WiMAX also introduce variables during devel- SymPA is a free software tool with packet- opment that developers working on fixed net- sniffing features, measurement localization works are unaccustomed to. Each technology maps, and other relevant functionalities (see

54 PERVASIVE computing Published by the IEEE CS n 1536-1268/10/$26.00 © 2010 IEEE

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. Related Work in Application Profiling

ymbian wireless network performance analysis tools, use, memory use, and key events. Mobile application eval­ S such as QualiPoc (www.swissqual.com), Nemo Handy uation requires not only using real devices but also testing (www.anite.com), and Customer Experience Manager (www. them in the field.2 SymPA addresses this key issue of mobile mformation.com), provide low-level information and focus on application development. specific service tests. However, unlike SymPA, they don’t let Finally, stability and security are crucial. Stability is key for developers debug their protocols’ behavior, solve connectivity mobile devices because they must operate for long time peri- problems, or correlate these issues with location data. ods. Users must always be able to switch away from any execut- Other solutions focus on providing on-target tests and de- ing application and make an emergency call. Symbian Signed bugging capabilities. JInjector is an instrumentation tool that (www.symbiansigned.com) provides a unified testing and cer- lets developers evaluate code coverage for Java 2 Micro Edition tification process to ensure that devices operate normally after applications both in emulators and on the actual devices.1 The applications are installed or removed and that applications are ability to debug applications on real devices is indispensable safe for users and don’t harm network operation. This initiative, because applications usually work successfully on the emulator as well as Java Verified (www.javaverified.com) and True Brew but fail on real targets. On-target testing is also one of the most Certification (http://brew.qualcomm.com), reveals that applica- time-consuming tasks and therefore requires simplification and tion integrity concerns not only end users but also developers, facilitation. mobile network operators, and phone manufacturers. Forum ’s Remote Device Access (RDA) service offers The mobile sector’s different stakeholders are clearly aware remote access to a wide range of mobile devices, so develop- that the correct performance of mobile applications is vital for ers can test their applications on a pool of real devices. RDA the technological evolution of devices, applications, and serv- also provides the possibility of testing on prototypes, which ices. But it’s also clear that communications performance, an lets developers design new applications that exploit these new open issue in application development,3 is beyond the scope of devices’ functionalities. Similar initiatives from Forum Nokia current development support tools. are the Virtual Developer Lab, powered by DeviceAnywhere, and the Device Loaner Program. These initiatives reveal the importance of using real devices during mobile application REFERENCES development. 1. M. Sama and J. Harty, “Using Code Instrumentation to Enhance Test- Deploying the application on real devices during testing is ing on J2ME: A Lesson Learned with JInjector,” Proc. 10th Workshop also necessary to analyze application performance on the hard- Mobile Computing Systems and Applications, ACM Press, 2009, pp. ware for which it was designed. The Nokia Energy Profiler lets 1–6.

developers monitor their applications’ energy use in real time 2. T. Halonen, J. Melero, and J.R. Garcia, GSM, GPRS and EDGE Perfor- and analyzes the processor, memory, and network signal levels. mance: Evolution toward 3G/UMTS, Halsted Press, 2002. The performance investigator delivered with Carbide.c++, an 3. K.A. Brown, “Impact of Wireless Communication on Multimedia Ap- Eclipse-based development tool supporting Symbian OS devel- plication Performance,” Multimedia Systems and Applications (Proc. opment, enables on-target data tracing of function calls, power SPIE, vol. 3528), 1999, pp. 566–573.

Figure 1) that let developers profile mo- and lets developers test applications methodologies, mobile device soft- bile applications with advanced com- programmed in different languages. ware development has a long time­ munication capacities. We developed University of Málaga researchers have scale.5 Networking is one of the most SymPA using Symbian, the most ex- successfully tested SymPA on projects important issues developers deal with tended OS for mobile phones, according ranging from mobile application devel- during implementation and testing on to the Canalys report “Smart Mobile opment to mobile service characteriza- real devices. Guaranteeing mobile ap- Device Shipments Hit 118 Million in tion over cellular networks.3,4 plication networking performance is 2007, up 53% on 2006” (www.canalys. key for the future of mobile devices. com/pr/2008/r2008021.htm). Symbian New Profiling Needs One reason this new generation of supports an ample range of program- in “Always Connected” mobile phones has been designated as ming languages, including Symbian Applications “smart” is their capacity to communi- C++, native C, native C++, Java, Py- Because programmers need to manage cate with many smart networks, rang- thon, Adobe Flash Lite, and Ruby, new APIs, OSs, and implementation ing from near-field communication to

JANUARY–MARCH 2010 PERVASIVE computing 55

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. PERVASIVE APPLICATION PROFILING

issue might seem trivial, but develop- ers must analyze more complex issues Traf c related to mobile communications’ capture changeable nature to discern between application performance issues or net- work errors. Battery Signal Developers should also be able to consumption strength cope with issues such as link outages. Connected However, to implement such function- mobile ality, they need tools to analyze traffic application pro ling and correlate it with location, speed, and wireless network issues. Improv- Data ing the application’s radio technology Mobility connection battery consumption is also a concern. QoS pro le The protocol stack configurations Localization on mobile devices differ from those (cell ID and of traditional PCs, so developers can’t GPS) use emulators. We don’t recommend using a mobile phone as a modem be- cause applications tested this way use Figure 1. Mobile application profiling. A complete characterization of mobile the computer protocol stack to estab- applications requires monitoring and correlating different types of information, lish data connections. So, developers such as IP traffic, received signal strength, location, and battery consumption. can’t use traditional protocol ana- lyzers to study networking perform- ance. To analyze connectivity issues, developers must execute the applica- tions on real devices, which requires tools that run on these devices. In ad- dition, adequate protocol behavior debugging tools might reduce devel- opment and testing time, but existing on-device debugging tools can detect only coding errors, not communica- tion underperformance. The SymPA profiling tool runs on real devices to help developers deal with these issues (see Figure 2). (a) (b) (c) SymPA also supports heterogene- ous environments. Heterogeneous Figure 2. SymPA (Symbian Protocol Analyzer) running on a Nokia 6110 Navigator. networks are one of the most interest- The mobile device provides advanced monitoring functionalities directly: (a) ing possibilities mobile devices offer, IP traffic capture, (b) quality-of-service (QoS) profile parameters, and (c) cell promising the development of “always information. connected” applications. However, heterogeneous environments are criti- cal scenarios in which connectivity global communication using cellular mobile communication performance could become a challenge for mobile technologies. errors. For example, implementing developers. SymPA enables perform- Wireless network and mobile phone servers on mobile devices is tedious ance evaluation of vertical handover networking issues require more effort because of limitations imposed by net- on mobile devices, which should al- than desktop computer issues because work operators. In this situation, de- low seamless roaming between dif- developers lack tools to check connec- velopers should be able to determine ferent networks. Developers could use tivity, analyze traffic data, study pro- whether a problem is related to closed SymPA to analyze these techniques’ tocol stack configuration, and debug ports or server implementation. This impact on IP communications from

56 PERVASIVE computing www.computer.org/pervasive

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. 160 500 0 2.5 General Packet Universal Mobile Radio Service Telecommunications 140 55 System 27 –20 400 2 120 70 –40 100 300 6 1.5 2 80 –60 11

Jitter (ms) 1 7 1 1 9 200 1 1 1 Consumption (W) Bandwidth (kbps) 60 1 1 2 –80 1 1 1 1 1 40 1 3 1 4 2 11 100 2 0.5 4 –100 Received signal strength indication (dBm) 20 5 1 15 1 1 2 22 1 2 0 0 2 1 –120 0 1 0:00 22 1 1:00 2:00 3:00 11 1 11 Time (min.:sec.) 1 1

Cell reselection Received signal strength Bandwidth Lost packet(s) Jitter Consumption Sequence error

Figure 3. Handover between General Packet Radio Service (GPRS) and Universal Mobile Telecommunications System (UMTS) networks. The handover causes a burst of packet losses, jitter increase, and bandwidth decrease. After handover, UMTS provides higher bandwidth, reducing the amount of lost packets at the expense of power consumption. the application’s viewpoint. For exam- lyzing this information is essential to cellular networks during the establish- ple, we’ve used SymPA to characterize identifying the source of connectivity ment of data connections. handover between GPRS and UMTS issues such as call drops, packet losses, Data traffic capture is fundamental networks. Figure 3 shows the burst or bandwidth reduction. SymPA helps for not only traffic analysis but also of lost packets during a handover on monitor all these parameters. debugging communication protocol a streaming session in a vehicle test. Figure 4 shows SymPA’s architec- implementations. SymPA users can It also shows each technology’s signal ture, including the libraries, APIs, and apply a wide range of measurements strength, jitter, bandwidth, and bat- different modules we implemented to to the established connections to ob- tery consumption. provide its analysis functionalities— tain parameters such as bandwidth, analyzing data connections, check- delay, and jitter. Other useful utili- Enabling Mobile ing connectivity, localizing measure- ties include incoming establishment- Application Analysis ments, tracking energy consumption, availability checking, which can deter- SymPA analysis centers on networking and correlating data. mine whether a mobile operator has issues. In general, radio propagation closed ports. conditions become unstable in wireless Analyzing Data Connections The mobile terminal’s memory scenarios, where factors such as signal SymPA’s data connection analysis stores the raw data traffic, which can quality and received signal strength— functionality includes traffic capture, be exported to libpcap. Because libp- usually affected by Doppler and received-signal-strength tracking, and cap is a widely used standard format, multi­path fading effects—trigger cell radio-access-technology identifica- it’s compatible with analysis tools such reselections and handovers between tion. SymPA also reports the quality- as Wireshark, a well-known protocol different access technologies. Ana- of-service (QoS) profile granted by analyzer.

JANUARY–MARCH 2010 PERVASIVE computing 57

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. PERVASIVE APPLICATION PROFILING

User interface Timer

Engine

Callback A SymP

QoSMet GPS Map Packet Connectivity Connection Signal Packet Data Battery interface location visualization capturing checking events events Protocol context consumption

Google Nokia Energy Pro­le

Static Third- Maps API external API party API s

HTTP engine

Sockets library

Localization Telephony Network interface High- library library resolution timer GPS Radio-access technology Battery Libraries and interfaces provided by the by provided interfaces and Libraries operating system (Series 60 Symbian OS) Symbian 60 (Series system operating

Figure 4. SymPA architecture splits functionalities into different modules. Some modules use third-party APIs, but the Symbian OS provides the main APIs.

Figure 5. Geographic localization Checking connectivity is particu- on a mobile device. SymPA provides larly useful in mobile-to-mobile use positioning features to locate cases because developers can’t deploy measurements using maps. traditional tools on the server side, as in a typical Internet scenario. SymPA incorporates a TCP client and a server, Checking Connectivity which allows file transfer between mo- SymPA lets developers verify mobile- bile devices. This active traffictraffi c generagenera-- to-Internet connectivity by establish- tion utility, in conjunction with the ing a data connection and then obtain- traffic capture feature, enables mobile- ing the IP address associated with the to-mobile connection performance connection. Obtaining the IP address evaluation. is useful because it changes each time A standard ping functionality lets a connection is established and can be users obtain an estimate of the con- used for further data processing or to nections’ round-trip delay. Develop- accept connections in servers running ers need a proper characterization of in mobile devices. communication delays to design timers

58 PERVASIVE computing www.computer.org/pervasive

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. and buffers adapted to the actual con- ditions that mobile applications will experience. Gps_SymPA.txt TIMESTAMP Latitude Longitude Speed 07/09/2008 11:21:40.7431 65.06216 25.48276 0.0000 Localizing Measurements 07/09/2008 11:21:42.7437 65.06215 25.48275 0.3454 0 7 / 0 9 / 2 0 0 8 1 1 : 2 1 : 4 4 . 7 4 1 7 6 5 . 0 6 2 51 2 5 . 4 8 2 7 5 0.4249 SymPA includes two features that NetworkInfo_SymPA.txt 0 7 / 0 9 / 2 0 0 8 1 1 : 2 1 : 4 6 . 7 4 1 1 6 5 . 0 6 2 51 2 5 . 4 8 2 7 4 0.1775 accurately locate the measurements TIMESTAMP RAT CELL ID BARS RSSI 7 collected and determine the associ- 07/09/2008 11: 11:20.8018 6 450800 7 89 07/09/2008 11: 11:21.9062 6 450800 7 89 ated speed for each piece of informa- 0 7 / 0 9 / 2 0 0 8 1 1 : 1 1 : 2 2 . 9 3 9 5 6 4 5 0 8 07 89 tion. First, the GPS-tracking function Trace_SymP0 7 / 0 A.txt9 / 2 0 0 8 1 1 : 1 1 : 2 3 . 9 6 8 7 6 4 5 0 8 07 89 07/09/200816:13:41.73510 7 / 0 9 / 2 0 0 8 1 1 : 1 1 : 2 5 . 0 0 0 0 6 4 5 0 8 07 89 activates monitoring of the mobile ..... 0 7 / 0 9 / 2 0 0 8 1 1 : 1 1 : 2 6 . 0 3 1 2 6 4 5 0 8 07 89 terminal’s speed and position. GPS------0 7 ---/ 0 9 */ 2* 0* 0CONTEXT 8 1 1 : 1 *1 *: 2* 7 . 0 6-- 2 5-- 6 4 5 0 8 0 0 7 Batter 8 9 yInfo_SymP A.txt related information, such as time Connection Speed5242888 1 1 : 1 1 : 2 8 . 0 9 T3imestamp 7 6 4 5 0 8 0 0 7 8 9 Batt level DataVolume Transferred8 1 1 : 1 1 : 2 9 . 3 1 4 0 6 4 5 0 07/09/20088 0 0 7 8 06:27:21.75479 100 stamps and coordinates, is stored in ...... 07/09/2008 11:11:26.8388 100 a file. Developers can use time stamps ------* * * Q o S * * * ------07/09/2008 14:31:54.8723 85 BER: 0.00001 07/09/2008 14:58:55.5921 71 during postprocessing to match dif- ...... 07/09/2008 15:26:56.2726 57 ferent information sources such as Max Rate Uplink: 384 07/09/2008 15:58:57.0210 42 Traf–c class: Background 07/09/2008 16:09:03.7520 42 captured traffic and radio condition ling priority: Unspeci–ed information. delay: 0 Second, the Google Static Maps API gives developers a map of the area directly on a mobile device (see Fig- ure 5). With this information, iden- Figure 6. SymPA profiling files use a text-readable format to store measurements tifying roads, buildings, and other and timing information. Postprocessing time stamps enable correlation of different objects that could affect communica- information sources, such as cell changes and packet losses. tion is easy. In addition, through post- processing of the SymPA-generated data files (see Figure 6), developers can create more meaningful and de- tailed maps, including coverage and handovers on a user’s path. Figure 7 shows how the GPS- tracking functionality’s location information allows a deeper un- derstanding of user mobility com- munication issues. To generate this representation, we used Google Earth, which integrates easily with SymPA. We tracked the received signal power of a user who was accessing an audio- streaming service while riding a bike from the VTT Technical Research Centre of Finland in Oulu to the city Figure 7. Correlation of Real-time Transport Protocol (RTP) traffic data, signal center. The blue vertical lines with strength, and cell data. The red line represents the received signal power during a numbered marks on top correspond commute. The blue vertical lines correspond to changes from one cell to another. to changes from one cell to another. The signal degrades around the change to cell 450787 and returns to acceptable During the experiment, degrada- levels on the other side of the river. tion in the sound quality and a tem- porary interruption occurred at some point near the river. After analyzing congestion, but rather to a significant tion occurs around the change to cell the captured information, we deter- loss of received signal strength at the 450787 and that the signal strength mined that the problems weren’t re- boundary between two cells. In the returns to acceptable levels on the lated to the server side or network figure, it’s evident that the degrada- other side of the river.

JANUARY–MARCH 2010 PERVASIVE computing 59

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. PERVASIVE APPLICATION PROFILING

120 200 0 Universal Mobile 269 Telecommunications System

100 –20 150

80 –40

60 100 –60 Jitter (ms) Bandwidth (kbps) 40 –80 50

20 1 –100 Received signal strength indication (dBm)

67 0 0 –120 0:00 1:00 2:00 Time (min.:sec.)

Cell reselection Sequence error Bandwidth Received signal strength Jitter Lost packet(s)

Figure 8. Data communication performance analysis on a mobile device. SymPA detects a link outage where the received signal falls for 20 seconds, interrupting the data connection. After 40 seconds, the communication continues in a new cell but loses 269 packets. In this way, SymPA helps identify the source of data connection errors.

Tracking Energy Consumption required in a UMTS connection could ciate them with locations. For exam- Because mobile devices are battery lead to confusion. This is because ple, while profiling a video-streaming driven, they require energy-saving the uplink power in UMTS might client over a public cellular network strategies.6 As a common practice, vary from insignificant values up to 2 in a vehicular scenario, we detected engineers need to pinpoint sources of watts, depending on the device’s ca- a streaming session that finished higher energy consumption and con- pabilities, the distance to the base sta- abruptly. So, we used SymPA informa- centrate later optimization on those tion, and the propagation conditions.7 tion to determine the problem’s source. critical functionalities. For that pur- Figure 3 shows how handover from Streaming quality depends on four pose, we added battery-monitoring GPRS to UMTS affects energy con- main factors: bandwidth, packet features to SymPA. sumption during a streaming session. loss, delay, and jitter. To calculate In particular, SymPA lets developers As expected, energy consumption is these factors, we analyzed Real-time fuse battery-consumption information higher in UMTS, and the consump- Transport Protocol (RTP) traffic cap- with all the other available informa- tion increase is more significant when tured during a streaming session with tion so that they can associate results the received signal strength decreases. Wireshark, then correlated this infor- with the actual conditions during the SymPA lets developers evaluate their mation with signal strength and cell measurement. Traditional profiling applications’ battery consumption us- data (see Figure 8). Signal strength de- tools obtain the energy-related infor- ing different radio-access technologies graded significantly below- 110 dBm, mation regardless of the spatial infor- in each location and context. a typical sensitivity threshold for mo- mation, which might negatively affect bile devices. results. Correlating Data The signal dropped for 20 seconds. For example, in a static environ- Correlating data from different levels This link outage caused the video- ment, comparing only the power re- lets developers find the source of er- streaming client’s buffer to underflow, quired in a WLAN connection to that rors at the application level and asso- and the application stopped executing.

60 PERVASIVE computing www.computer.org/pervasive

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply. the AUTHORS

Adding location information to this Almudena Díaz is a researcher at the University of Málaga. Her research inter- correlation process, we determined ests are communication protocols, wireless networks, and mobile application that this error occurred when the ve- development. Díaz has an MS in telecommunication engineering from the University of Málaga. She’s a member of the IEEE and ACM. Contact her at hicle was in a tunnel. [email protected].

n the past few years, we’ve been

involved in multiple projects and Pedro Merino is an associate professor at the University of Málaga. His re- have obtained a firsthand view search interests are foundations, tools, and applications of formal methods of the issues and challenges that for critical systems, particularly communications software and development Imobile application developers must techniques for Internet services on mobile networks. Merino has a PhD in computer science from the University of Málaga. He’s a member of the IEEE face with networking analysis, de- Communications Society and ERCIM. Contact him at [email protected]. bugging, and profiling mechanisms. Today, there are smarter devices for which we need smarter tools and so- F. Javier Rivas is a PhD candidate at the University of Málaga. His research in- lutions. Our ongoing work aims to terests are mobile computing, software verification, and wireless communica- simplify development of mobile appli- tions. Rivas has an MS in telecommunication engineering from the University of Málaga. Contact him at [email protected]. cations by providing communication awareness so that applications can exploit all the intelligence available in these new smarter phones. We’re working to provide an application-level framework for isolat- ing networking issues, which develop- Information Networking and Applica- ers can incorporate in their applica- tions (AINA 07), IEEE CS Press, 2007, pp. 784–791. tions. We’re also integrating SymPA 2 Free and QoSMeT,8 a passive measure- 4. A. Diaz et al., “Evaluating Video Stream- ment tool for measuring one-way end- ing over GPRS/UMTS Networks: A Prac- Sample Issues! tical Case,” Proc. IEEE 65th Vehicular to-end network QoS developed by the Technology Conf. (VTC 07), IEEE Press, VTT Technical Research Centre. In- 2007, pp. 624–628. tegrating these tools will enable one- 5. D. Wood, Symbian for Software Lead- way mobile device measurements in ers: Principles of Successful real time. Development Projects, Symbian Press, 2006.

6. T. Zhang et al., “Reducing Energy Con- ACKNOWLEDGMENTS sumption on Mobile Devices with WiFi Interfaces,” Proc. Global Telecommuni- The Spanish government-sponsored project TIN cations Conf. (Globecom 05), IEEE Press, 2008-05932 and the Autonomous Community 2005, vol. 1, pp. 561–565. of Andalusia project P07-TIC-03131 funded this research. 7. User Equipment (UE) Radio Transmis- sion and Reception (FDD), tech. specifi- The magazine of cation TS 25.101, �����������������������3rd Generation Partner- computational tools ship Project (3GPP), 2008. REFERENCES and methods for 1. F.H.P. Fitzek and F. Reichert, Mobile 8. J. Prokkola et al., “Measuring WCDMA 21st century science. Phone Programming and Its Application and HSDPA Delay Characteristics with to Wireless Networking, Springer, 2007. QoSMeT,” Proc. IEEE Int’l Conf. Com- munications (ICC 07), IEEE Press, 2007, pp. 492–498. 2. A. Ocampo et al., “Toward a Reference Send an e-mail to [email protected] to Process for Developing Wireless Internet receive the two most recent issues of CiSE. Services,” IEEE Trans. Software Eng., (Please include your mailing address.) vol. 29, no. 12, pp. 1122–1134.

3. A. Diaz et al., “Experimental Analysis Selected CS articles and columns http://cise.aip.org of Peer-to-Peer Streaming in Cellular are also available for free at www.computer.org/cise Networks,” Proc. Int’l Conf. Advanced http://ComputingNow.computer.org.

JANUARY–MARCH 2010 PERVASIVE computing 61

Authorized licensed use limited to: University of Texas at Arlington. Downloaded on April 24,2010 at 22:41:01 UTC from IEEE Xplore. Restrictions apply.