[IEEE 22nd International Conference on Data Engineering (ICDE'06) - Atlanta, GA, USA (2006.04.3-2006.04.7)] 22nd International Conference on Data Engineering (ICDE'06) - Monitoring Top-k Query inWireless Sensor Networks

Download [IEEE 22nd International Conference on Data Engineering (ICDE'06) - Atlanta, GA, USA (2006.04.3-2006.04.7)] 22nd International Conference on Data Engineering (ICDE'06) - Monitoring Top-k Query inWireless Sensor Networks

Post on 11-Apr-2017

212 views

Category:

Documents

0 download

Embed Size (px)

TRANSCRIPT

<ul><li><p>Monitoring Top-k Query in Wireless Sensor Networks</p><p>Minji Wu Jianliang Xu</p><p>Hong Kong Baptist Univ.Kowloon Tong, Hong Kongxujl@comp.hkbu.edu.hk</p><p>Xueyan TangNanyang Technological Univ.</p><p>Singaporeasxytang@ntu.edu.sg</p><p>Wang-Chien Lee</p><p>Penn State Univ.University Park, PAwlee@cse.psu.edu</p><p>Abstract</p><p>Top-k monitoring is important to many wireless sensorapplications. This paper exploits the semantics of top-kquery and proposes a novel energy-efficient monitoring ap-proach, called FILA. The basic idea is to install a filter ateach sensor node to suppress unnecessary sensor updates.The correctness of the top-k result is ensured if all sensornodes perform updates according to their filters. We showvia simulation that FILA outperforms the existing TAG-based approach by an order of magnitude.</p><p>1 Introduction</p><p>Owing to the rapid advances in sensing and wire-less communication technologies, wireless sensor networkshave been available for use in a wide range of in-situ sensingapplications, such as habitat monitoring, wild-fire preven-tion, and environmental monitoring [4]. A wireless sensornetwork typically consists of a base station and a group ofsensor nodes (see Figure 1). The base station serves as agateway for the sensor network to exchange data with ex-ternal users. The sensor nodes, on the other hand, are re-sponsible for sensing and collecting data from their localenvironments. They are also capable of processing senseddata and communicating with their neighbors and the basestation.</p><p>Monitoring of aggregate functions is important to manysensor applications and has drawn a lot of research atten-tion [3, 4]. Among those aggregates, a top-k query continu-ously retrieves the set of k sensor nodes with the highest (orlowest) readings [1, 2]. However, how to energy-efficientlyanswer top-k queries is a great challenge to wireless sen-sor networks. The sensor nodes usually operate in an unat-tended manner and are battery powered; replacing the bat-</p><p>Minji Wu and Jianliang Xu were supported in part by Research GrantsCouncil of Hong Kong under Project No. HKBU 2115/05E.</p><p>Wang-Chien Lee was supported in part by National Science Founda-tion grant IIS-0328881.</p><p>Wireless Sensor NetworkBase Station</p><p>Sensor Updates</p><p>Filter Updates/ Probe</p><p>User</p><p>Result</p><p>Query</p><p>Figure 1. The System Architecture</p><p>teries is not only costly but also impossible in many situ-ations (e.g., in a hard-to-reach area). If a certain portionof the nodes run out of their power and lose their coverage,the whole network will be down. Thus, in addition to reduc-ing network traffic, a distinguished requirement for wirelesssensor networks is to balance the energy consumption at thesensor nodes to prolong network lifetime.</p><p>A basic implementation of monitoring top-k querywould be to use a centralized approach where all sensorreadings are collected by the base station, which then com-putes the top-k result set. In order to reduce network traf-fic for data collection, an in-network data aggregation tech-nique, known as TAG, has been proposed [3]. Specifically,a routing tree rooted at the base station is first establishedand the data is then aggregated and collected along the wayto the base station through the routing tree. Consider a sim-ple example shown in Figure 2a, where sensor nodes A, B,and C form a routing tree. The readings of these sensornodes at three successive sampling instances are shown inthe tables of Figure 2a. Suppose we are monitoring a top-1query. Employing TAG, at each sampling instance, nodesB and C send their current readings to the parent (i.e., nodeA), which aggregates the data received with its own read-ing and sends the highest (i.e., the readings from node C inthis example) to the base station. The top-1 result is alwaysnode C, but nine update messages (three at each samplinginstance) are used. As such, this approach incurs unneces-sary updates in the network and, hence, is not energy effi-cient.</p><p>In this paper, we exploit the semantics of top-k query</p><p>1</p><p>Proceedings of the 22nd International Conference on Data Engineering (ICDE06) 8-7695-2570-9/06 $20.00 2006 IEEE </p></li><li><p>#1 35#2 38#3 37</p><p>525651</p><p>B C</p><p>484543</p><p>#2 56#3 52</p><p>#1 51#1 43</p><p>#3 48#2 45</p><p>525651</p><p>Base Station</p><p>A</p><p>(a) TAG</p><p>#1 35#2 38#3 37</p><p>#2 56#3 52</p><p>#1 51#1 43</p><p>#3 48#2 45</p><p>48</p><p>48</p><p>52</p><p>52</p><p>Base Station</p><p>A</p><p>CB</p><p>[20, 39)</p><p>[47, 80)[39, 47)</p><p>probe node C</p><p>1</p><p>1</p><p>2</p><p>3</p><p>3</p><p>(b) FILA</p><p>Figure 2. An Example of Top-k Monitoring</p><p>and propose a novel filter based monitoring approach calledFILA. The basic idea is to install a filter at each sensor nodeto suppress unnecessary sensor updates. The base stationalso keeps a copy of the filter setting to maintain a view ofeach nodes reading. A sensor node reports the reading up-date to the base station only when it passes the filter. Thecorrectness of the top-k result is ensured if all sensor nodesperform updates according to their filters. Figure 2b showsan example, where the base station has collected the initialsensor readings and installed three filters [20, 39), [39, 47),and [47, 80) at sensor nodes A, B, and C, respectively. Atsampling instances 1 and 2, no updates are reported since allupdates are filtered out by the nodes respective filters. Atinstance 3, the updated reading of node B (i.e., 48) passesits filter [39, 47). Hence, node B sends the reading 48 tothe base station via node A (step ). Since 48 lies in thefiltering window of node C (i.e., [47, 80)), the top-1 re-sult becomes undecided as either node B or C can have thehighest reading. In this case, we probe node C for its cur-rent reading to resolve the ambiguity (steps and). Thus,a total of four update messages and one probe message areincurred in this approach.1 Compared with the aforemen-tioned TAG-based aggregation approach, five update mes-sages are saved at the cost of one probe message. Obvi-ously, this approach achieves a better performance than theTAG approach.</p><p>Yet, in order to make FILA to work efficiently, two fun-damental issues arising at the base station server have to beaddressed:</p><p> How to set the filter for each sensor node in a coordi-nated manner such that the top-k result set is correctlyreturned if all nodes perform updates according to theirfilters? The filter setting is critical to the performanceof FILA. In the above example, if nodes B and C havethe filters set to [39, 50) and [50, 80), respectively, no</p><p>1For simplicity, the overhead for initial data collection and filter settingis not shown here, but counted in our experiments.</p><p>updates need to be reported for all three samplings.</p><p> Upon receiving an update from a sensor node, how toreevaluate the top-k result and how to update the af-fected filters?</p><p>We answer in this paper the above two questions with theobjective of reducing network traffic and prolonging net-work lifetime.</p><p>2 System Model and Problem Definition</p><p>We consider a wireless sensor network as depicted inFigure 1. It is assumed that the base station has continu-ous power supply and its radio strength is strong enough tocover all sensor nodes. In other words, a probe messagebroadcast by the base station can reach all sensor nodes ina single hop. In contrast, the sensor nodes are poweredby battery. Their radio coverage is constrained to a localarea. When the base station is beyond a sensor nodes radiocoverage, an underlying routing infrastructure (e.g., a TAGtree [3]) is used to route data to the base station.</p><p>Each sensor node i measures the local physical phe-nomenon vi (e.g., pollution index, temperature, or residualenergy, etc.) at a fixed sampling rate. Without loss of gen-erality, we consider a top-k monitoring query that continu-ously retrieves the (ordered) set of sensor nodes R with thehighest readings, i.e.,</p><p>R =&lt; n1, n2, , nk &gt;,</p><p>where i &gt; j, vni vnj and l = ni(i =1, 2, , k), vl vnk . The monitoring result is maintainedby the base station and updated to the user. To producecontinuous query results, the proposed monitoring approachcontrols when and how to collect sensor reading updates tothe base station.</p><p>3 FILA Overview</p><p>Initially, the base station collects the readings from allsensors. It then sorts the sensor readings and obtains theinitial top-k result set. Next, the base station computes afilter (represented by a window of [li, ui)) for each sensornode i and sends it to the node for installation. At the nextsensor sampling instance, if the new reading of sensor nodei is within [li, ui), no update to the base station is needed.Otherwise, if the new reading goes beyond the filtering win-dow and passes the filter, meaning the top-k order might beviolated, an update is sent to the base station. The base sta-tion will then reevaluate the top-k result and adjust the filtersetting(s) for some sensor node(s) if necessary. The queryreevaluation algorithm is discussed in detail in [6].</p><p>2</p><p>Proceedings of the 22nd International Conference on Data Engineering (ICDE06) 8-7695-2570-9/06 $20.00 2006 IEEE </p></li><li><p>As can be seen, the purpose of using filters is to filter outsome local sensor updates and hence suppressing the trafficin the network. The correctness of the top-k result must beguaranteed provided that all sensor nodes perform updatesaccording to their filters. Thus, the filter settings have tobe carefully planned in a coordinated manner. Denote thecurrent reading of node i by vi. Without loss of generality,we number the sensor nodes in decreasing order of theirsensor readings, i.e., v1 &gt; v2 &gt; &gt; vN , where N is thenumber of sensor nodes under monitoring. Intuitively, tomaintain the monitoring correctness, the filters assigned tothe nodes in the top-k result set should cover their currentreadings but not overlap with each other. On the other hand,the nodes in the non-top-k set could share the same filtersetting. Thus, we consider the filter settings only for thetop-k+1 nodes. A feasible filter setting scheme, representedas {[li, ui) | i = 1, , k + 1}, must satisfy the followingconditions:</p><p>u1 &gt; v1;vi+1 &lt; ui+1 li vi, (1 i k);lk+1 vN .</p><p>(1)</p><p>u2= l13</p><p>v12vv3vv5</p><p>=l2uu =u45 =l3l5=l4</p><p>4</p><p>descending order</p><p>u1</p><p>Figure 3. Filter Settings for Top-3 Monitoring</p><p>Figure 3 shows a feasible filter setting for top-3 monitor-ing, where nodes 4 and 5 share a filter setting and ui+1 isset equal to li for 1 i 3 in order to maximize the filter-ing capability. Intuitively, a filter setting is a (constrained)partitioning of the data space. A straightforward way is toset the filter bound at the midpoint of two sensor readings,i.e.:</p><p>ui+1 = li =vi + vi+1</p><p>2, (1 i k). (2)</p><p>We call it uniform filter setting. It is favorable in the casewhere the sensor readings from all sensor nodes follow asimilar changing pattern. A more sophisticated scheme forfilter setting is described in [6].</p><p>We have developed a simulator based on ns-2 and NRLssensor network extension to evaluate the proposed FILA ap-proach. Figure 4 shows the results against TAG [3] for awide range of real traces [5]. We can see that FILA im-proves network lifetime over TAG by an order of magnitudewhile achieving a much lower average energy consumption.</p><p>4 Conclusions</p><p>This paper proposed a novel energy-efficient approachcalled FILA for top-k monitoring in wireless sensor net-</p><p>(a) Network lifetime</p><p>(b) Average energy consumption</p><p>Figure 4. Performance Comparison with TAG</p><p>works. As for future work, we plan to extend the proposedmonitoring approach to other aggregate functions such askNN, average, and sum. We are going to build a proto-type based on Motes and measure the performance in realenvironments. We are also interested in monitoring spatialqueries in object-tracking sensor networks.</p><p>References</p><p>[1] B. Babcock and C. Olston. Distributed top-k monitoring. InProc. ACM SIGMOD, pages 2839, June 2003.</p><p>[2] P. Cao and Z. Wang. Efficient top-k query calculation indistributed networks. In Proc. PODC, July 2004.</p><p>[3] S. Madden, M. J. Franklin, J. M. Hellerstein, and W. Hong.TAG: A tiny aggregation service for ad-hoc sensor networks.In Proc. USENIX OSDI, pages 131146, December 2002.</p><p>[4] R. Szewczyk, E. Osterweil, J. Polastre, M. Hamilton,A. Mainwaring, and D. Estrin. Habitat monitoring with sen-sor networks. Communications of the ACM, 47(6):3440,June 2004.</p><p>[5] Tropical Atmosphere Ocean (TAO) Project.http://www.pmel.noaa.gov/tao/data deliv.</p><p>[6] M. Wu, J. Xu, X. Tang, and W.-C. Lee. Top-k monitoringin wireless sensor networks. Technical Report, Hong KongBaptist University, June 2005.</p><p>3</p><p>Proceedings of the 22nd International Conference on Data Engineering (ICDE06) 8-7695-2570-9/06 $20.00 2006 IEEE </p><p> /ColorImageDict &gt; /JPEG2000ColorACSImageDict &gt; /JPEG2000ColorImageDict &gt; /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 2.00333 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict &gt; /GrayImageDict &gt; /JPEG2000GrayACSImageDict &gt; /JPEG2000GrayImageDict &gt; /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.00167 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict &gt; /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /False</p><p> /Description &gt;&gt;&gt; setdistillerparams&gt; setpagedevice</p></li></ul>

Recommended

View more >