New User Special Price Expires in

Let's log you in.

Sign in with Facebook


Don't have a StudySoup account? Create one here!


Create a StudySoup account

Be part of our community, it's free to join!

Sign up with Facebook


Create your account
By creating an account you agree to StudySoup's terms and conditions and privacy policy

Already have a StudySoup account? Login here

Sensor Networks

by: Buck Ankunding

Sensor Networks CMPE 259

Buck Ankunding
GPA 3.56

Katia Obraczka

Almost Ready


These notes were just uploaded, and will be ready to view shortly.

Purchase these notes here, or revisit this page.

Either way, we'll remind you when they're ready :)

Preview These Notes for FREE

Get a free preview of these Notes, just enter your email below.

Unlock Preview
Unlock Preview

Preview these materials now for free

Why put in your email? Get access to more of this material and other relevant free materials for your school

View Preview

About this Document

Katia Obraczka
Class Notes
25 ?




Popular in Course

Popular in Computer Engineering

This 180 page Class Notes was uploaded by Buck Ankunding on Monday September 7, 2015. The Class Notes belongs to CMPE 259 at University of California - Santa Cruz taught by Katia Obraczka in Fall. Since its upload, it has received 59 views. For similar materials see /class/182218/cmpe-259-university-of-california-santa-cruz in Computer Engineering at University of California - Santa Cruz.


Reviews for Sensor Networks


Report this Material


What is Karma?


Karma is the currency of StudySoup.

You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!

Date Created: 09/07/15
CMTE 259 Sensor Networ s Katia OEraczKa Winter 2005 Transport frotocoLs II 11 lnnouneements CI feed aekon project proposals CI Trq eet resources 12 Transport protocols cont if D KMS I Cl C OQL ZL CI Summary 13 EMS CI Keh a e Muti5egment Transport CI W ere to do re a ih ty 0 914546 0 Transport 0 lpp cation 15 M ZLC relia ilitt CI 8 0211 C RTSCTS 19am ick O QBasio stopandwait MQ 0 Mo MwaLen in Broadcast or mutica5t modes 0 Random 50t seection CI Options 0 N0 MQ O fQ ways O 5dective MQ 16 M lC relquiliti cont d CI Witnout MQ 0 lee Eroudeust mode 0 For unieast uddress screening at routing dyer O s no over eud I With MQ O Zlnieast transmissions O For Eroud e r mutieust use mutipe unieust O Num er ofretries is configuru e CI 5eeetioe MQ O Zlnieast uses MQ O roud und mutieast use no MQ 39 3 route discovery 17 Transport relialiiliti CI 5triety e2e O I nitiatecf 5y sink CI Loedreeovery O I ntermecfiate nodes trigger repair when 055 is detected 0 Modes eae e packets CI N CKEased 18 pplieationauer reliaEilitt CI ireetecfdt usion Eased O 5ink5emfs out request quotinterest O W en e0mpete data received sinkremoves request 19 Question CI enefits of owerayer re a ih ty Cl cfcfitiond over ead 110 K9145 overview CI functions 0 fragmentationreassem j 0 Guaranteed Je very CI Unique identifiers O Nofragments 0 fragment id s and numEer offragments CI Loss detection and repair 0 Sequence FLoes and timers 0 Loss detection at either sinks or intermediate nodes 0 N CKs 111 freliminarq analysis Cl Demonstrate t e Eenefits of opEy op re a i ty 112 KMST evaluation CI WConQJ re a tfity CI Load recovery 0 Wit and without M C re a i ty CI Endtoemfre a i ty O Wit and without M C re a i ty 113 OEservations CI W en there is no transport reia5iity 0 MC reiaEiity eritiea in ossy inks CI HopEynop transport reia5iity O ldds itte to reiaEe MC 0 QBut opEy op transport reiaEiity ony more e ieient t an adding MC reiaEiity 39 MC liRQooernead incurred in every packet CI EZE transport reia5iity O W en no MC reiaEiity is used simuation does not terminate opEy op recovery is eritieaL O fa121C reiaEiity used opEy op and e2e transport reiaEiity are equivaknt 114 OEseroations cont a Cl Experiments witn ig error rates 0 HopEynop transport reia5iity without M lC reia5iity O HopEynop transport reia6iity5e MQ O 2e transport reia6iity 5e MQ Cl Hen transport reia5iity witnout MQEreaamps down at ig error rates 0 Routing has hard time estaEisFLing routes 115 COD 4 116 Canestion etection and lvoicfance Cl Importance of congestion contro 117 W at is C 0234 P Cl Energy efficient congestion eontro Cl anee mee nnisms are invoked O Congestion detection 0 Open00p opEy op Eackpressnre O Cosecfoop mntisonree regniation 118 Congestion detection Cl eearate and efficient congestion detection is important 0 Channe oacfing 5ampe channe at appropriate rate to detect congestion 119 Open00p Eu Eackpressare Upstream node decides to propagate Eaekpressare or not 120 Closed 0019 multisource regulation 121 Congestion detection scnemes CI affer occupancy 0 Not reia6e in CSM networks Cl Cnanne oadinCa O Goodfor the immediate neig or ood C Energy considerations CI Report rate 0 Report rate goes down congestion O etection Based on report rate needs to react on onger time scae 122 COD 21 overview CI ComEinntion of Enckpressnre fast time scne witn dosed oop congestion contro Cl nckpressnre targets quotoch congestion w ereas dosed oop regiintion targets persistent congestion Cl nckpressnre is c enpersimper since it s open oop CI Congestion contro requires nfeecfonckhop 0 Uses 521C K om sink to se dock 123 C OQL ZL performance metrics CI verage Energy Tax 2 Tota oackets dropped in network Tota oackets received at sink Cl verage fide ty Tena Ey 2 Difference Eetween average nnmEer of packets deioered at sink using C 02991 and using idea congestion 5cne1ne 124 Simulation Setup Cl Random networktopdogies wit networksize om 30 to 120 nodes CI 2M5p5 IEEE 80211 me RESC75 are amend Cl Directed diffusion is used as routing core Cl fixed work 0ad 6 sources and 3 sin s Cl 5oaree generate data at different rates Cl Event packet is 64 Eytes and interest packet is 36 Eytes 125 5imuation Remit Case 1 QDense Source 5qu Kate I I Fa21E1 Iellaer Lrae CE I E31 auntarewary tracEnaE n pulsararr l r 53 H Em aamemdrzuptracanazzsz a Number u FaI ie a m 4 I TIITIE E 5imuation Remit Case 2 Sparse Sources Low Rate g I 39 I I quott 39h 39quotrvl 391 n Ek 3 Ei hEbj I n39hdei 39er5 JuCZE El Ii sun but u l39 lquotuii39 b 127 5imuation Results Case 2 Sparse Source Low Rate 33 11 l 15E DUE MunEwJI I Leia 128 5imuation Remit Case 3 Sparse Sources 5qu Kate 2D I I I I Hamnzc ty l H l Lti39i39rgHEEIEIEEAETaa I I I Iogu39ILCIDidEETEI n E 1E 1W39D ETE rn 4U 5539 E31 E E1 125 Network Size no of nodes 129 Conclusion Cl COD 21 5 energy efficiency Cl COD 21 s n5iity t0 nan e persistent and transient congestion 130 TTCI SpOTt protocok summary 131 El For sinktosource communication 23 network reprogramming El Ke a i ty via retransmissions El Sequencedriven 055 Success Rate v Channel sum rule page toss m Figure 2 Probability of successful delivery of a message JEtECtiUTL when the mechanism allows multiple retransmissions before the next packet arrival CY Wan AT Campbell and L Krishnamurthy PSFQ A Reliable Transport Protocol for Wireless Sensor 132 Networks WSNA39OZ September 28 2002 Atlanta Georgia USA W CI Endtoemf or hophyhop repair the atter is genera y Better Cl Suggests that repair coahf he done at either M C ayer MQ retransmissions or Transport Layer requests hasecf on agment namhers etc CI imercfriven oss detection amt oca cfata caches Cl fits with the irectecf i hasion ET F Stann and J Heidemann RMST Reliable Data Transport in Sensor Networks IEEE SNPA O3 133 2 Cl imfar avera guafity afserm ce rather than nadeitainade re a ifity 16 7 14 Opumalaperaung pcml P 1f A 12 7 5 mqmm wn 21 g m 7 a a g u a R A E g g g u a 7 E I cm M g u 2 r w m 1Uquot 1U Sensor Networks quot In Proc ACM IobiHoc 0 u Repurlmg frequency 0 Sankarasubramaniam Y Akan OB and Akyildiz IF quotESRT EVenttoSink Reliable Transport in Wireless 3 m CI Receiver Easedcongestion detection CI Open 00p apEy ap Eackpressure CI CwseJLaop mutisource regu ztion mH39nD mom gum Figure 7 Experimental Eur network Lemma mpuxogy Nudge are well cmmecced Packets we unjnasL 39 Y Akan n R and Alzvildi T F quotESR I 39 quot L 39 39anspm t in Wireless Sensor Networksquot In HULACMMUbIHac Q CMPE 259 Sensor NeTwor39ks Ka ria Obr39aczka Winter 2005 Storage and Querying II Announcements Cl Hw3 is up D Exams El Signup for project presentations Cl Schedule Cl Course evaluation Mon 0314 0 Need volunteer Today Cl Storage D Querying Da ra Cen rric S romge DCS Cl Da ra dissemination for39 sensor ne rwor39ks Cl Namingbased s ror39age Background El Sensorne r o A dis rribu red sensing ne rwork comprised of a large number of small sensing devices equipped wi rh processor memory radio 9 Large volume of da ra Cl Da ra Dissemination Algori rhm o Scalable o Selforganizing 9 Energy efficien r Some definiTions Cl ObservaTion o Lowlevel oquuT from sensors 9 Eg deTaiIed TemperaTure and pressure readings Cl EvenT o ConsTeIIaTions of lowlevel observaTions o Eg elephanTsighTing fire inTruder Cl Query 9 Used To eliciT The evenT informaTion from sensorneTs 9 Eg is There an inTruder Where is The fire Da ra dissemina rion scheme Cl Ex rer39nal S ror39age ES Cl Local S ror39age LS Cl Da raCen rr39ic S ror39age DCS Ex rernal S romge ES Local S rom 396 LS 1 10 Local S romge LS DaTa CenTric STorage DCS Cl EvenTs are named wiTh keys Cl DCS provides key value pair39 Cl DCS suppor39Ts Two oper39aTions 0 PuT k v sTor39es v The observed daTa according To The key k The name of The daTa E Gel k r39eTr39ieves whaTever39 value is sTor39ed associaTed wiTh 3Y El Hash funcTion a Hash a key k inTo geographic coor39dinaTes o PuT and GeT oper39aTions on The same key k hash kTo The same ocaTion 112 DCS Example O Put eephant data 11 28 M I O 1128Hash elephantquot O O 113 DCS Example I Get eephant o 39 1 1 28 I FDA l 1128Hash elephant O O O Q O O 0 114 DCS Example con rd Q C 1 o O O 39 39 39 Geographic Hash Table GHT Cl Builds on a PeerTopeer39 Lookup Sys rems Q Greedy Perime rer39 S ra reless Routing GHT Peertopeer GPSR lookup system 116 Comparison study Cl Me rr39i cs 0 To ral Messages 9 To ral packe rs sen r in The sensor ne rwork o Ho rspo r Messages 4 Maximal number39 of packeTs senT by any par ricular39 node 117 Comparison s rudy con r39d CI Assume o n is The number of nodes 6 Asymp ro ric cos rs of 007 for floods 007 12 for39 poin r ropoin r rouTing ES LS DS CosT for39 S ror39age 007 12 0 00712 Cos r for Query 0 007 00712 Cos r for Response 0 00712 00712 118 Comparison Sngy conT39d D 0 The ToTaI number of evenTs deTecTed Q The number of evenT Types queries for39 04 The number39 of deTecTed evenTs of evenT Types DD D No more Than one query for39 each evenT Type so There are Q quer39Ies In ToTal D Assume hoTspoT occurs on packeTs sending To The access ponnT 119 Comparison S rudy con r39d ES LS DCS TOTO39 DumpE QnDq71 Q7 Dmmz Dq QMDWM Q Summary HoTspoT D total 2 Q summary 9 DCS is preferable if I Sensor network is large I Dtotal gtgt maXan 120 Summarx CI In DCS relevanT daTa are sTored by name aT nodes wiThin The sensorneTs El GHT hashes a key k inTo geographic coordinaTes The keyvalue pair is sTored aT a node in The ViciniTy of The locaTion To which ITs key hashes El To ensure robusTness and scalabiliTy DCS uses PerimeTer Refresh ProTocol PRP and STrucTured ReplicaTion SR El Compared wiTh ES and LS DCS is preferable in large sensorneT 121 Mul ri Resolu rion S romge Goals Cl Provide s rorage and search for raw sensor data in da rain rensive scien rific opera rions Cl Previous work 0 Aggrega rion and querying 0 Focus on applications whose i n reres rs are known a priori 123 Approach Cl Lossy progressively degrading s rorage 124 Cons rrgc ring The hierarchy Jr W39x nape r n m haw Cons rrgc ring The hierarchy amrg l Processing of each level Store incoming summaries locally forfuture search Get compressed Decode i summaries from Reencpde at lower chiIdren a X y resolution and fonNard to parent 127 Cons rruc ring The hierarchy Recursiver send data to higher levels of the hierarchy Dis rribu ring s romge load i a Em 1 29 What happens when storage fills up9 Cl Eventually all available storage gets filled and we have to decide when and how to drop summaries Local Storage Allocation Res 4 Res 3 Res 2 Res 1 Local storage capacity CI Allocate storage to each resolution and use each allocated storage block as a circular buffer 130 Tr39odeoff between storage requirements and query quality Storage time El Graceful query low query accuracy l low 0 degradafon providing high mPa tness 39 5 more accurate r39espo nses to queries on recent data and less accurate responses To f g queries on older data highquotlt high query accuracy low compactness How to allocate storage at each node to summaries at different resolutions to provide gracefully degrading storage and search capability 131 Match system performance to user39 requirements User provides a function Quser that r 95 represents desired query quality degradation over time System provides a step function ststem with steps at times when sWF 39 39 are aged Query Accuracy 50 past Time present Objective Minimize worsf case difference befween userdesired query quaiTy bue curve and query quaiTy Thar The sysfem can provide red step function 132 For how long should summarie 3e s rored CI To achieve desired query quali ry given system39s cons rrain rs Cl Given 0 Nsensor nodes 0 Each node has s rorage capaci ry 5 0 Users ask a se r of Typical queries 739 O Da ra is genera red a r resolu rion i a r ra re R O Dqk query error when drilldown for query q Terminafes a r level k 0 QM Userdesired quali ry degradation 133 Aging strategy with limited information full a priori information Omniscient Strategy baseline when entire data is available Training Strategy when small training da et from initial deployment Greedy Strategy when no data is available use a simple weighted allocation to summaries 1 34 No a Distributed tracedriven implementation Cl Linux implementation 0 Uses Emstar J Elson et al a Linuxbased emulatorsimulator for sensor networks 0 3D Wavelet codec 0 Query processing CI Geospatial precipitation dataset 0 15x12 grid 50km edge of precipitation data from 19491994 from Pacific Northwest Cl System parameters 0 Compression ratio 6212224248 0 Training set 6 of total dataset 135 How efficient is Search Error vs Level 07 I I A GlobalDailyMax Query 8 GlobalYearlyMax Query 00 0396 LocalYearlyMean Query E D 05 P S 04 I U lt1 3 03 E E 02 C 9 8 01 0 I I I Level in hierarchy where query terminates Search is very ef cient lt5 of network queried and accurate for different queries studied 136 Comparing aging strategies 1 I 5 DinTriispipni rHll39llllt 09 39il Grc dy Detail 39 a quot Greecy Balanced E 18 ill Greedy Duraliun i g I T i 1 IE 13 i 39 a J J 39I l E 391quot 39 quotquot I E 39 ME 5 3 K S 1 L H 12 39 a LL k k A M J Ll l39fi mm 39 D I I I I U 5 ID 15 EU 25 erzal Skmrage Size MB Training performs Within 1 to optimal Careful selection of parameters for the greedy algorithm can provide surprisingly good results Within 2 5 of optimal 137 Summarx El Progressive aging of Summaries can be used To supporT long Term spaTioTemporal queries in resourceconsTrained sensor neTwork deploymenTs D We describe Two algoriThms a Trainingbased algoriThm ThaT relies on The availabiliTy of Training daTaseTs and a greedy algoriThm can be used in The absence of such daTa El Our resuTs show ThaT 0 Training performs close To opTimaI for The daTaseT ThaT we sTudy O The glreedy algoriThm performs well for a wellchosen summary weig Ting parameTer 138 Con rinuously Adaptive Con rinuous Queries CACQ CACQ In rroduc rion CI Proposed con rinuous query CQ sys rems are based on s ra ric plans 0 Bu r CQs are long running 0 Ini rially valid assump rions less so over Time El CACQ insigh r apply con rinuous adap rivi ry 0 Dynamic opera ror ordering 0 Process mul riple queries simultaneously 0 Enables sharing of work amp s rorage 1 4O Ou rline CI Background 0 Mo riva rion O Con rinuous Queries 0 Eddies Cl CACQ O Con rribu rions Example driven explana rion Cl Resul rs amp Experiments 141 Mo riva ring applica rions Cl Building moni roring Cl Varie ry of sensors eg ligh r Temperature vibra rion s rrain e rc Cl Varie ry of users wi rh differen r in reres rs eg s rruc rural engineers building managers building users e rc 1 42 Confinuous queries Cl Long running s ronding queriesquot 0 From various users 0 On a number of sensor s rreoms El Ins rolled continuoust produce resul rs un ril removed Cl LOTS of queries over The same do ro sources 0 Oppor runi ry for work sharing 1 43 Eddies amp adap rivi ry Cl Eddies Avnur amp Hellers rein SIGMOD 2000 Con rinuous Adap livi ry D No s ra ric ordering of opera rors Cl Rou ring policy dynamically orders opera rors on a per ruple basis Cl done and ready bi rs encode where ruple has been where if can go 144 CACQ con rribu rions El Adop rivi ry El Tuple lineage 0 In oddi rion ro ready and done encode po rh ruple Takes Through opero ror Enables sharing of work and s ro re across queries Cl Grouped fil rer O Efficiently compu re selec rions over mul riple queries Cl Join shoring Through S ro re Modules S reMs 1 45 Eddies amp CACQ Single Query Single Source Use ready bi rs To Track wha r To do nex r All 139s in single source Use done bi rs To Track wha r has been done Tuple can be ou rpu r when all bi rs se r Rou ring policy dynamically orders ruples SELECT FROM R WHERE Ra gt10 AND Rb lt15 m g A T4 139 v I V 1quotquot E l Gill 1 46 Evalua rion Cl Real Java implementation on Top of Telegraph QP 0 4000 new lines of code in 75000 line codebase Cl Server Pla rform 0 Linux 2410 0 PenTium III 733 756 MB RAM Cl Queries posed from separa re works ra rion O Ou rpu r suppressed CI Lo rs of experimen rs in paper jus r a few here 1 47 CMPE 259 Sensor Ne rwor39ks Ka ria Obr39aczka Winter 2005 Localization GPS Annogncemen rs Cl Homework due on 0214 0 Submission email To ka ria cin ria 0 Plain rex r or39 pdf Cl Final project pr39esen ra rions CI Upda red reading lis r for nex r class Today Cl Gabriel Elkaim39s Talk on GPS Node Lgcaliza rion Node Localiza rion D For some sensor ne rwork applica rions exac r loca rion is cri rical 0 Tracking 0 Moni roring D For mos r applica rions having loca rion informa rion enhance value of informa rion CI Also needed in geographic rou ring CM PE 259 Sensor NeTworks KaTia Obraczka WinTer 2005 TransporT ProTocoIs AnnouncemenTs El ProjecTs posTed DSome projecTs will be presenTeddiscussed aT The end of class Today DProposals due by Friday 0121 m WhaT is expecTed ouT of a TransporT proTocol for sensor neTworks ReliabiliTy congesTion conTrol Why can39T we use The exisTing proTocols Resource consTrainTs power sTorage compuTaTion complexiTv daTa raTes Mo riva rion con1quotd El Applica rion specific El Spec rra for known cons rrain rs Mo riva rion con1quotd In general SWSP El Simple Wireless Sensor Pro rocol El Design challenges 0 Limited capabilifies DAssumpTions o Fixed nefworkquot Topology 0 Access poinfs as dafa collecfors Why no i39 TCP El Too heavy duTy DCongesTion conTrol and wireless links 0 Disable congestion confro 0 Low bandwidfh El Buffer size 0 Small windows El Mul riple connec rions 0 Single connecfion SWSP overview W Observations El Sensor registers with an AP 0 Listens for RR messages 0 Sends registration 0 Waits for ACK gt connected state El Window size El Periodic KA from sensors El Data retransmitted after 3 retries El ACKS piggybacked onto RR messages El Data piggybacked onto KA messages W El Methodology 0 Platform C with Linux Simulated different sensors as different processes AP simulated using another PC 39 Wireless communication 0 Metrics Throughput of bytes received by APtime Delay timeACKrecd timedatasent SWSP evaluaTion lconT d El ThroughpuT increases up To cerTain number of sensors Then decreases as sink geTs overrun El Delay increases subsTanTially beyond a given number of sensors DSoluTions EvenT To Sink Reliable TransgorT ESRT for Wireless Sensor NeTworks SalienT FeaTures El EvenT To sink reliabiliTy El Self adjusTing CI Energy awareness low power consumpTion requiremenTl El CongesTion conTrol El DifferenT compleXiTy aT source and sink ESRT39s definiTion of reliabiliTy El ReliabiliTy is measured in Terms of The number of packeTs received Or re orTing frequency ie number of packeTsdecision inTerva El Observed reliabiliTy number of received daTa packeTs in decision inTerval aT The sink El Desired reliabiliTy number of packeTs required for reliable evenT deTecTion El Reporfing raTe number of packeTs senT by sensor over Time inTerval El Normalized reliabiliTy observeddesired ESRT Qroblem definiTion DeTermine reporTing frequency of source nodes To achieve required reliabiliTy aT sink wiTh minimum resource consumpTion Preliminary observations El Reliability increases as reporting frequency increases up to a certain threshold El Why ESRT operation Algorithm for ESRT El If congestion and low reliability decrease reporti g frequency aggressively exponential decrease El If congestion and high reliability decrease e rting to relieve congestion No com promise on reliability multiplicative increase El If no congestion and low reliability increase reportin frequency aggressivelymultiplicative increase El If no congestion and high reliability decrease reporting slowing half the slope Components of ESRT El In sink 0 Normalixed reliability computation o Congestion detection mechanism El In source 0 Listen to sink broadcast 0 Overhead free local congestion detection mechanism Eg buffer level monitoring 6N Congestion Notification Performance results gbased on simulations CI Starting with no congestion and low reliability i 539 39 Performance results cont d CI Starting with no congestion and high reliability Performance results cont d CI Starting with congestion and high reliability Performance results cont d CI Starting with congestion and low reliability Performance resulTs conT39d El Average power consumpTion while sTarTing wiTh no congesTion and high reliabiliTy Challenges wiTh ESRT El MulTiple concurrenT evenTs DIS There a way To slow down The nodes causing The congesTion El OThers PS FQ m El MosT sensor neTwork applicaTions do noT need reliabiliTy 0 Sources gt sink El New applicaTions like re Tasking of sensors need reliable TransporT o Sink gt sources El CurrenT sensor neTworks are applicaTion specific and opTimized for ThaT purpose El FuTure sensor neTworks ma be general purpose To some egtltTenT a iliTy To re program funcTionaliTy Goals Probabili ry of successful delive using end To end mo el El Simplicify El Robusfness DScalabiliTy M 17p El Cusfomizabilify l 11 11quot plslhcecmmlc ufwlrelesslmk m belweenlvm hups w Goals of PSFQ Pumg Slowly and Fe rch Qulcklx DUDE uecoven from losses locally Minimum signalin perale correctly in Iossy environments Independent of underlying routing infrastructure Mulf39 hop packe r forwarding When no link Loss 7 multihap farwmding lakexplare Recovering from errors How PSF recovers from errors s rore and forward quot Error recovery memagex are wasted No wane of error recovery memagex PS FQ ogera l ion El Al rerna re beTween mulTi hop forwarding when low error ra res and s rore and forward when error ra res are higher El 3 func rions 0 Pump message relaying 0 Error recovery fetch o Sfufus reporfing report PSFQ Pumg Schedule If not duplicate and inorder and TTL not 0 then Cache and schedule for forwarding at time t TmmlttltTmax Ms Fe l39ch Quicklyquot Ogera l ion When loss detected then fetch mode Loss aggregation try to recover a window of lost packets Proac l39ive Fe l39ch quot Regor l El ReporT aggrega rion DCarries s ra rus informa rion node id seq El Triggered by user 0 Injecf dafa message will reporf bif sef Performance evaluation DCompare wi rh SRM Scalable Reliable Mul ricas r El Performance Me rrics 0 Average Delivery Ratio 0 Average Lafency 0 Average Delivery Overhead Experimen ral se l39up Er39r39or39 Tolerance in YulImlcv m m E E ag EEE EI 3 M 2 Mbps CSMACA Channel Access Tm IOOms Tmm SOms Tr20ms m m m min Overhead Average la l39encx Lamncy vs a lnnol mm 11 CMPE 259 Sensor Networks Kcrria Obraczka Winfer 2005 Roufing Profocols II Announcemen rs El Reading assignment 1 is up No res on Dir ec red Diffusion El Mul riple PCI H39IS can be used To forward da ra back To The sink El Is if The same as mul ricas39r W Targe r degloymen rs El Sparse ne iworks El MulTi Tiered deployments o Sensors 0 Wired access poinfs o Mules Apgroach El Mobile agen is El MULEs mobile ubiqui rous LAN extensions 0 Mobiliiy oCornrnunicafionslior139 range uws radios low power and abiiiiy ia handle bursts o Buffering Pros and cons Pros and cons El Pros 0 Energy efficiency 39 Listen for the mule o InferrniH39enf connecfivify El Cons 0 Increased lafency Al rer na rives Approaches Latency Power Reliability Infrastructure cost Base stations Low High High High Adrhoc Medium ML Medium MH MULE High Low Medium Low 3fier ar chi rec rur e El Wired APs El Mules El Sensors Considera r ions El APs have no limiTaTions El Mules o Sfor age mobilify abilify To communicafe wifh sensors and A s o Unpredicfable movemenf paH39er39ns 0 Can Talk To ofher mules Benefits El Robustness El Reliabili ry More considera rions El No rou ring overhead El Mules can Transport daTa for mul riple applications El High IaTency 0 Delay bounds El Mobili ry limiTaTions Sys rem model El Simple discreTe El LoTs of assumptions 0 Realisfic El Performance meTrics o Reliabilify o Buffer size 0 Delay Main resul rs El Buffer requiremenTs aT sensors inversely proporTional To raTio of number of mules To gri size El Buffer requiremenT aT mule inversely proporTional To raTio of number of mules To grid size and raTio of APsTo grid size El Relationship beTween buffer capaci ry number of mules and reliabiliTy Energyefficien r rou ring Schurgers e r al El Two approaches 0 Efficienf dafa collecfion using aggregafion 0 Load balancing spread Traffic uniformly ObservaTions El Energy animal rouTing needs To consider fuTure Traffic 0 Energy limiTaTions Tn A and E send 50 pm To B B T Fsends 100 pm ma Load balancingADB ECBFDB c D But if nodes can only send 100 pm D would no be able m deliver all of F39s pm to B E A F 1mm caseACB ECB FDB EnergyefficienT versus ener o Timal El STaTisTically opTimal and only considers ausal informaTion El LifeTimeworsT mse Time unTil node fails Traffic spreading El Make sure ThaT nodes are used uniformly by rouTing El GradienT based rouTing GER o DirecTeddiffusion varianT 0 Use shorTesT paTh in number of hops To sink To forward daTa El Performance meTric E o RooT mean square of The PDF of energy used by nodes Traffic sgreading approaches El STochasTic node picks new hop randomly chosen from neighbors wiTh equal gradienT El Energy hnsed node increases iTs quotheighTquot when iTs energy falls below a cerTain Threshold All nodes need To adjusT Their heighT accordingly El STream bsed diverT sTreams from nodes ThaT are parT of paThs used b oTher sTreams ResulTs El TargeT Tracking scenario El STream timed spreading performs The besT El STochasTic spreading does beTTer Than energ timed and pure GBR Krishnamachari eT al EnergyrobusTness Tradeoff in mulTigaTh rouTing El MulTipaTh for robusTness o FaulTTolerance Through redundancy El AlTernaTively reduce number of i mediaTe no es 0 Single paThs o Nodes use higher TransmiT power ConsideraT ions El Energy meTric number of Transmissions TransmiT ener o IndependenT of number of receivers El RobusTness meTric o ProbabiliTy message reaches sink in The face of node ailures 0 Assume ThaT nodes fail wiTh probabiliTy p independenle from oTher no es El PareTo opTimaliTy criTeria o RouTing scheme dominaTes anoTher iff more robusT wiTh sTrichy less energy or o Iff iT uses equal or less energy wiTh sTrichy higher robusTness CMPE 259 Sensor szwor39ks Kutiu Obruczku W Wzr 2005 Rommg Announcements Transgor f Ero l ocols summa Pump Slow Fefch Quickly PSFQ n For smkrtor reprogrammmg u Rehabmry m ztransm SS OHS D quuz czrdr vz n3mEm itJAZT ZSJJZJJCCZ SSN oss dzvzcnon unnlmxhrnrrh 9 5m H RMST w n Am fur uvemH quamy ufserv 2 mm Man nudemrnude rehab ny W h u Endrvorznd or hoprbyrhop repaw m2 aHzr s generaHy bznzr u Suggests mm repaw coma be done m 2mm MAC mm ARQ retransmwssmns or Transpom Layer requests based on fragment mum em a Txmzrrdmvzn oss dmmon and oca dam caches D FMS WW1 the Dv emed D ffus on API bers 17mm mmmrmwknmmmnsmmmsmwz 6 w m A R M Summarlzmg Transgor f Issues El Receiver based congestion detection D Because of harsh condmons and severe constramts H may be bznzr m mwemem rehabe m ahoprbyrhop D OPEquot l P MP by MP buckpress39fre rather than andrrwnd manner a mm m MAC or u ClosedLoop mulhsource regulation quotmpgquot my a For energy ef meno n s bzsr to avmd congzsnon zmwzxy or have packet osses occur dose m W Source Back pressure 5 auszm techmque u Where posswb e schedu zd so unons are preferab e w m my mum 1 u Issueschalleng es ay special aTTenTion To any 2 El DifficulT To p d RouTing individual no 0 CollecTing information wiThin The specified region El Sensors may be inaccessible 0 Embedded in physical sTrucTures o Thrown inTo inhospiTable Terrain More issueschallenges er and environmental demands also 39cs39 More issueschallenges El Topological issues 0 ArbiTrarily large scale 0 FrequenT Topology changes Bu sTion AccidenTs New nodes are added El Us conTribLITe To dyn ml 0 Nodes move 0 ObjecTs move El DaTacenTric and applicaTioncenTric view 0 LocaTion 39me 0 Type of sensor 0 Range of values More issueschallenges El Not nodetonode packet switching but node tonode data propaga Ion El High level tasks are needed Is it the time to order more inventory Challenges El Energylimited nodes El Computation 0 Aggregate data 0 Suppress redundant routing information El Communication 0 Bandwidthlimited o Energyintensive Challenges El Scalability adhoc deployment in large scale 0 Fully distributed wo global knowledge 0 Large numbers of sources and sinks El Robustness unexpected sensor node failures El Dynamics no aprl39arl39 knowledge 0 Sink mobility 0 Target mobility Directed Diffusion A Scalable and Robust Communication Paradigm for Sensor Networks C Intanagonwiwat R Govindan D Estrin Application Exampl Remote Surveillance O 39Ei39ve me periodic repar fs abaLf ani39ma aea an In region A every f seconds 0 Te me in Whaf dire an fliaf vehi39ce in region y 5 ma Vi39ng Basic Idea El Iririetwork data processing eg aggregation cac 39 El Distributed algorithms using localized interactions El Applicationaware communication primitives o Expressed in terms of named data Elements of Directed Diffusion ii NamIi o Dafais named using attributevalue pairs El Inferesrs o A node requests data by sending interests tornarned ata El gradient 0 Sradienfsis set up within the network designed to draw39 events ie data matching the interest ii Reinforcement o Sink reinforces part cuiar neigh en bars to draw higher quality higher data rate ev ts El Content based namin Tasks are named by a list of attr bute vuue pa the attributes 0 An mul tracking Requesr Reply Interest Task Description Type towniegged anirn i Interva 20 ms Duration 1 minute Location a100 7100 200 400 a Time 02 10 35 0 Task descr39 ption specifies an interest for data matching Node data Type founiegged animai instance iepnant Interest CI The sink periodica y braadcasfs interest messages to each o its neigh ors El Every node maintains an interest cache 0 Each item corresponds to a distinct interest 0 No information about the si k 0 Interest aggregation identical type completely overlap rectangle attributes El Each entry in the cache has several fields 0 Timestamp last received matching interest 0 Several gradients data rate duration direction Setting Up Gradient m l SoLirce e Neighbor s choices 1 Floodln 2 Geogaphlcroutlng 3 Cache olata to direct interests Interest Interrogation Gradient o is interested data rate duration direction Data Propagation El Sensor node computes the highest requested event rate among all its outgoing gradients El When a node receives data 0 Find a matching interest entry in its cache Examine the gradient list send out data by rate OCache keeps track of recent seen data items loop prevention 0 Data message is unicast individually to the relevant neighbors The nelg1bor reinforces a path 1 Al least one neighbor K tom whom latestevent low delay 39 k hlch 2 Choose the onef lt rsl receivedLhe 3 Choose all nelgh orsfromw ents were recently received LOW rate event gt Reinforcement Increased interesl m Local Behavior Choices I For propagating interests I In the exam e flood I More sophisticated behaviors possible e g based on cached information 6P5 I For setting up gradients I dafa rafe gradienfs are sef up fawards neighbors w 0 send an inferes I Others possible probabilistic gradients energy gradients etc Local Behavior Choices El For data transmission 0 Mulripa r delivery with selective quality along different paths 0 Probabilistic forwarding o Singlepath delivery etc El For reinforcement o Reinforce parrs based on observed delays 0 Losses variances etc Initial simulation study of diffusion El Key metric 0 Average Dissipated Energy per event delivered ind cutes energy eff ciency and network lifet me El Compare diffusion to o Flooding o Centrally computed tree omniscient multicast Diffusion Simulation Details El Simulator Its2 CI Network Size 50250 Nodes El Transmission Ran e 40m El Constant Density 195x103 nodesm2 9 8 nodes in ms a El MAC Modified Contentionbased MAC El Ener y Model Mimic a realistic sensor radio Pottie 20009 0 660 mW in transmission 395 mW in reception and 35 mw In idle Diffusion SimulaTion El Surveillance applicaTion rces are randomly selecTed wiThin a 70m x 70m corner in The field 0 5 sinks are randomly selecTed across The field 0 High daTa raTe is 2 evenTs sec 0 Low daTa raTe is 002 evenTssec o EvenT size 64 byTes o InTeresT size 36 byTes o All sources send The same locaTion esTimaTe for base experimenTs Average DissipaTed Energv g quotquot15 nndiny Ea nma 5 3 nmz g 5 nm 3 nnnn amniscieni Mullicasl g g nnns a z nmiisinn 3E mum g 5 nnnz o 4 391 39 5n 1 15n Znn 25n 3 n Network Slle Diffusion can outperform ooding and even omniscient multicast suppiess duplicate location estimates Conclusions I Can leverage daTa I Achieve desired global behavior Through I Empirically adapT To observed environmenT Energy fficienT mulTigaTh rouTing Energyefficien r mul riga rh rou ring El Based on direcfed diffusion El In direcfed diffusion o Sink broadcasfs inferesf o Sensors periodically low rafe sends back dafa e g evenf defecfion reporfs o Sink sends reinforcemenf on preferred pafh 0 Reverse pafh is esfablishe 0 Upon missing reporfs sink rebroadcasfs inferesf and sink reinforces Problem El Periodic flooding of interests and events in The presence of failures El Solufion Solu rion mul rigle 120th El Mulfipa fh roufing 0 Load balancing o Reliable delivery by sending duplicafes o Robusfness Observa rions El Primary pa fh best path El Dafa sent at lower rate on alferna fe pa fhs El Upon failure on primary pa fh reinforcemenf on alternate pa fh El If all alfremafe pa fhs fail flooding for path re esfa 39 hmenf El Overhead alfernafe pa fh mainfenance El Resilience measured as how often pa fh re esfablishmenf is neede Approach El Disjoin f versus quotbraidedquot pa fhs El How To build multiple paths wifh local information only Localized disioin r mul riga rhs El Sink esfablishes primary pa fh El Sink selecfs next bestquot neighbor quotAquot El A propaga fes alferna fe afh39 reinforcemenf To ifs besf neighbor quotBquot El If B is already on a path between sink and source B sen s ack a negative reinforcemenfquot El Access To local information only may lead To longer pafhs Braided mul riga rh El Parfially disjoin f El For each node on primary pa fh find best path from source To sink That does not contain Thaf node El Pa fhs in The braid expend equivalen f energy El Reinforcemenf To best node and alfernafe reinforcemen f To next bestquot node m El Energy efficiency 0 Overhead El Resilience To failures 0 Isolafed versus paH39erned failures ResuH s n Braided multipatiis are more energy efficient 0 Especially al iower densities at Disjoint multipatiis have better resilience to patterned losses d multipatiis exiiibit better resilience to isolated failures Geograghic roufing u Deliver packets to nodes or regions based on their geographic location a Typically nodes know their position and immediate neighbors Basic Geog raghic Forwarding u Greedy send packet to neighbor that is closest i to desiinuf on a Can get stuck in voids GPSR proposes a perimeter routing mode to avoid this a no in has our ammo mks menkssuaw s madam Trajecfo Based Forwarding u preeencode arbllrar geographic lranclor y39 packet goes inroagn nodes ciosesl l0 lhls lranclory a particularly well sailed for large networks Wiln m n ensii g y y quot4 p o Mama thymswdvuwmgmmmmm mecummm Geographic rouTing wiThouT locaTion informaTion Rao eT al El Apply geographic roLITing when mosT nodes do noT have posiTion informaTion El Approach vrfLa canrdnafesquot 0 Use local connecTiviTy informaTion AssumgTions El Nodes know Their own coordinaTes El Nodes know coordinaTes of nodes in The 2 hop neighborhood RouTing El Greedy forward To neighbor closesT To desTinaTion El When packeT arrived To desTinaTion sTop El If sTLIck do expanding ring search LInTil closer node found CoordinaTe consTrLIcTion El A node39s coordinaTes is The average of iTs neighbors39 coordinaTes El Finding perimeTer nodes39 coordinaTes o Beacon nodes flood quotHelloquot message 0 PerimeTer nodes discover disTance in hops To oTher perimeTer no es 0 PerimeTer nodes broadcasT Their perimeTer vecTor o PerimeTer nodes use TriangulaTion To find coordinaTes of all perimeTer nodes Energy ConsumpTion Issues in Sensor Ne rworks Cin ria B Margi CMPE259 March O9Th 2005 Ou rline Energy Model for Communica rions MASCOTSO4 paper Energy consump rion for Processing Tasks Power TOSSIM SenSysO4 PredicTionbased energy map Adhoc Journal 05 Energy Harves ring ISLPED39O3 SoE UCSC Energy model for39 communica rion MASCOTS 04 paper by Cin ria amp Ka ria SoE UCSC Energy model for communico rion Powerawareness in sensor ne rworks MAC proTocols SMAC YeOZ TRAMA Rojendron03 TMAC vonDomO3 Direc red Diffusion Infonogonwiwo rOO oggregoTion SolisO4 QuolNe r GloMoSim and nsZ EiTher do no r model all The radio s ro res Or do no r Toke proper occoun ring Accoun ring done on differen r layers SoE UCSC Energy model for communication Relo red Work Measuremen rs of energy consumed by NICs NICs in hondhelds STemm97 WoveLAN Iop rops FeeneyOl Models LEACH HeinzelmonOO Sensor neTwork life rime BhorwojOZ Measure bo r rery discharge To model communico rions Lochin03 SoE UCSC Energy model for39 communication FeoTures Explicile accoun rs for39 lowpower r39adio modes Considers The differen r energy cos rs ossocio red wi rh each one of The possible r39odio s ro res For39 example State TR1000 WaveLAN Transmitting 2475 mW 1400 mW Rxoverhearingsensing 135 mW 900 mW Idle 135 mW 900 mW Sleeping 15 uW SoE UCSC Energy model for39 communication Model Energy spen r while in a given r39adio s ra re y y P V i Y y TX Ty PackefSizeTransmissionRafe OTher39wise use a Timer39 Implemen red in GloMoSim and QualNe r SoE UCSC Energy model for communication Valida rion Sani ry check compare wi rh original GloMoSim Tes rbed in SMAC paper More on MASCOTSO4 paper SoE UCSC Energy model for communication Valida rion IEEE 80211 Original vs InsTrumen red GloMoSim Simula rion parame rers Node 0 Nodel Node 2 No mobili ry CBR Traffic node 0 To 2 data size is 200 bytes Dura rion is 250 seconds Energy parameters for radio original GIoMoSim SuE UCSC Energy model for39 communication ValidaTion SMAC Quoli ro rive comparison Simulo rion vs Tes rbed Cl C SMAC r39o rocol YeOZ p ode 2 0 N d 0 Nada 3 5node 2hop Topology App 10 x 380 by res Low power39 r39odio TRIOOO SimuloTionmeosur39emen rs Ios rs enough Time for39 all packe rs To be Transmi r red 10 SoE UCSC Nod 1 N d 4 Energy model for communication ValidaTion SMAC Same behavior as 4 r esul rs in Ye02 Source average P nodes 0 amp 1 Message mevrAmva mm s 50E UCSC Case S rudies Pr39o rocol comparison 80211 vs SMAC MASCOTS 2004 Analy rical Model Valida rion Singlehop sa rur39a red IEEE 80211 wireless neTwork ICCCN 2004 12 SoE UCSC Energy model for communicafion 80211 vs SMAC Parame l39er39s 50 nodes low power r adio TRIOOO CBR wi rh 1o sources 380 bytes 3 r ou ring AODV Dum rion 1505 per node N nsummmn Message mevrAmva mm s SoE UCSC Energy model for communica rion 80211 vs SMAC Average Time per state IAT 1s 1 000 100 I 80211 I SMAC 01 TX RX Overhear Sensing Idle Sleep 14 SoE UCSC Energy model for communication Summary Simple energy model for communico rion Implemen red a r GloMoSim amp QuolNe r Ins rrumen ro rion provides comple re energy and Time accoun ring per radio s ra re Useful Tool To evolua re and unders rond poweraware pro rocols 15 SoE UCSC Processingsensing energy model ongoing work SoE UCSC 16 Processingsensing energy model SoE UCSC For39 simple sensor39s eg Tempero rur39e energy consumed by communico rion subsys rem domino res However for39 more sophis rico red sensor39s eg occelerome rer39s amp mogne rome rer39s This is no r Tr39ue Doher39TyOl How obou r camera as sensors 17 Processingsensing energy model Relo red Work Energy savings due do ra compression Borr03 Power manogemen r archi rec rure for Iap rops BalakrishnanOl Power Managemen r in Wireless Ne rworks ZhengO3 Energy budge r Greo r Duck Island deploymen r MoinworingOZ 18 SoE UCSC Processingsensing energy model Approach Energy cos r based on Tasks Energy measuremen rs Curren r Discharge raTe 19 SoE UCSC Processingsensing energy model Tes rbeds Dell Iap rops S rar39ga res Mo res SoE UCSC 20 Processingsensing energy model MeThodology Macroscopic view Se r of experimen rs baseline sys rem processing FFT disk access dbench for IapTops ne rwork Transmission Iperf for lapTops Ne rwork recep rion Iperf for Iap rops Wellknown benchmarks whenever possible 21 SoE UCSC Processingsensing energy model Me rhodology Lap rops Power Management off Use ACPI To ob rain vol rage amp discharge raTe S randard for power managemen r Define meThods To read The parameTers Under Linux procacpi Every rime a file in procacpi is read corresponding ACPI me rhod is execuTed 22 SoE UCSC Processingsensing energy model MeThodology S rarga res amp Mo res STargaTes measure curren r using power suply use ba r rery moni ror chip Vladi39s projec r AAoTes measure curren r using power suply Sami r39s projec r wi rh mo res 23 SoE UCSC Processingsensing energy model Resul rs Laptops Task Av Discharge Rate Baseline 10200 W FFT 25047 W Disk 13430 W TX 22389 W RX 16101 W SoE UCSC Stargates Task Current Idle 475 mA FFT 735mA TX 740mA RX 700mA sleep 67mA 24 Then who r From a comple re energy consumption choroc rerizo rion we can derive energy consump rion predic rion model application dependen r hardware dependen r resource manager 25 SoE UCSC Smart usage of energy in sensor nodes Transmit full stream Or a bit notifying Define a methodology 3 for sensor nodes to 2 make decisions that allow energy savings Interesting application Visual Sensor Nodes 50E 7 chc Power TOSSIM SenSysO4 SoE UCSC Power TOSSIM SenSysO4 ex rension To TOSSIM TinyOS Simula ror To include energy consump rion add a module Tha r keeps Track of power sTaTe modifica rions To o rher modules To repor r Transi rions CPU energy usage gt es rima re number of cycles in AVR genera re Traces Tha r will processed la rer28 SoE UCSC Power39 TOSSIM MicoZ Power39 Model Mode Current Mode Current CPU Radio Active 80 mA Rx 703 mA Idle 32 mA Tx power 00 372 mA ADC Noise Reduction 10 mA Tx power 01 521 mA Powerdown 103 pA Tx power 03 537 mA Powersave 110 pA Tx power 06 647 mA Standby 216 pA Tx power 09 705 mA Extended Standby 223 pA Tx power 0F 847 mA Internal Oscillator 093 mA Tx power 60 1157 mA Leds 22 mAled Tx power 80 1377 mA Micaz sensorboard 07 mA Tx power C0 1737 mA EEPROM Tx power FF 2148 mA Read 62 mA Read time 565 ps Write 184 mA Write time 129 ms SoE UCSC Power TOSSIM BenChmGVd her benchmarks Energy mJ 2750 u Beacon Ehnk CntTur CntTo Oscmu Osclunr Sense TmyDE TmyDB Surge Leas LeasAnd scope scopeRF Light me select Rfm ToLog ngm PowerTOSSM is accurate less than 15 error for all tesis Graph from slides used at SenSysO4 by authors 30 SnE ucsc Predic rion based energy map Ad hoc Journal 05 SoE UCSC Predic rion based energy map Ad hoc Journal 05 Goal cons rruc r an energy map of a wireless sensor neTwork using predic rionbased approach Naive approach nodes send periodically upda res wi rh i rs available energy To moni roring node Problem 32 SoE UCSC Predictionbased energy map Approach Nodes send a message with cur39r39en r energy available and par39ame rer39s of energy dissipa rion model Nodes send upda res if pr39edic rion is off by a pr39ede rer39mine Threshold eg 3 33 SoE UCSC PredicTionbased energy map Energy dissipaTion model ProbabilisTic model based on Markov chains node operaTion modes are The sTaTes TransiTion probabiIiTy maTrix is consTrucTed based on The node pasT hisTory Then can calculaTe energy dissipaTed based on Time spenT on each sTaTe 34 SoE UCSC Predictionbased energy map Diagram mumm Flg 2 Diugmm or the smterbused energy dissipation modele SoE ucsc Energy Harves ring ISLPED39OS SoE UCSC Energy HarvesTing ISLPED39OS HarvesTing problem problem of exTracTing The maximum work ouT of a given energy environmenT Goal learn abouT energy environmenT energy available and recharging capabiliTies use This info for Task sharing among nodes 37 SoE UCSC Energy Harves ng ChaHenges workload X recharging cycles residual energy is no r enough info so need To know how recharging occurs needs To predic r recharging oppor runi ries o rherwise consider only residual energy 38 SoE UCSC Energy Harvesfing EEHF algori rhms Energy SoE UCSC Mausu FEIHL39HI ut Eu vim u mmlml NI IUE l S39pchral Estimation Energy Cmneumplluu FredicLion n E h Filler E m l 1quot Iquot in 39 Sltxlmmic E C39mmum l39 Em 39 11 mm I Predicbm Sculuhi i gr Friend y Nclwurk In fur mat in n ExulmugL39 Elwrgy Awan d Sclmduliug TupuluY v M gull ling x J Figure l Interactiunoi between mr39mus ICICliIIquotalg n139thma Energy sources Tibia 1 A comparison M e nargvsnums Elem mums Power enam mm mumps media WEE d m39 htm mums malarng Li T iurm l m mme 3 41 W 39 hr W 15 mWa omi39 om an n CI 1 summ1 l39cioudy ag So a MIME DEE mm zm39w gamma o11rze mash 5prt1Wa cmi lt MW desk Himm quotVimam BASH 3911 mWJ39om39i mm 3EIE mm39a am i39 31 35 Du iiiE 41 mWa c m quot at 1123 Em Pasmn humm11 f t aysmms 13 mw she2 Insensa Hunter Maw 30 r21 l i l39nrr13 139 SE mWn Jamquot Edna are sari mama taken from Im mmue1 nnnlyquotim And a tagw experinmnm quotWellies am highly dependent mm the amplirr Lde39 and 39 Icequcncy39 m the cl r Wing Uil39lmrmm Paper PicoRadio Supports Ad Hoc UltraLow Power Wireless Networking 40 SoE UCSC Energy sources Microbial Fuel Cells EcoBoT II hTTpwwwiasuweacuk Anode bacTeria found in sludge acT as caTalysTs To generaTe energy from The given subsTraTe flies or roTTen apple CaThode 02 from free air acTs as The oxidising agenT To Take up The elecTrons and proTons To produce HZO EcoBoT I Anode a freshly grown culTure of E coli fed wiTh refined sugar CaTholyTe ferricyanide 41 SoE UCSC Ener39gy sources Microbial Fuel Cells MFC X Alkaline ba r rer39y SoE UCSC single MFC oquuT vol rage is 08V capaci ry is 163mAh and energy is 37mWh IT weighs 1009 and cosTs 300 AA alkaline cell oquuT vol rage of 15V capaci ry of 28Ah and an energy is 42Wh IT weighs 259 and cos rs 030 42 CMPE 259 Sensor Ne rwor39ks Ka ria Obr39aczka Winter 2005 Storage and Querying Overview Overview Cl Sensors sensegenero re do ro CI Usersapplications in reres red in do ro CI Require on infrastructure for do ro access and s roroge Cl Common user opero rions are 0 Queries monitoring 0 Ac ruo re and con rrol Types of queries Cl Hisfor39icali O Wha r is The average rainfall over39 pas r 2 days Cl Current 0 Wha r is The current Temperate in Rm 226 CI Long running 0 Temperature in r39m 226 over39 The nex r 4 hours every 30 seconds Issues D How To iden rify relevon r sensors CI Compu ro rion vs communication rrodeoff 0 Where To process query Inside The sensor neTwork rouTe query AT cenTrolized locoTion rou re doTa Large amounts of data transfer ef ciency 0 How To process query DaTaSDace Queryinq and Monitorincl Deeply NeTwor39ked CollecTions in Physical SDace T Imielinski and S Goel RuTger39s UniversiTv Cl Billions of objec rs popula re space Cl Each produces and locally s ror39es da ra Cl Loca rion awar39e Cl Can be selec rively moni ror39ed quer39ied and controlled 9 Physical wor39ld enhanced with da ra Charac reris rics Cl Da raspace O Da ra lives on The objec r 0 Users access no r only local i nforma rion bu r can naviga re en ri re da raspace O Spa rial world divided in 3D dafacubes CS Bldg sTreeT block eTc O Communica rion messaging and compu ra rion Techniques for querying and moni roring required Qgerying and moni roring Cl Queries are spa rially driven Cl S l39eps O Iden rify relevan r da racubes O Iden rify relevan r nodes da raflocks DaTacube direc rory service 0 Aggrega rion for queries on several da racubes Eginforma rion abouT ManhaTTan Taxi cabs Archi rec ring Da raSpace Cl Querying and moni roring O Mul ricas r mechanisms aTTribuTe value pair mapped To a mul ricas r address 0 Group membership based on Physical locaTion ATTribuTe Tempera rure vehicles eTc 0 Mapping a r rribu re value To mul ricas r address Eg using hash Tables 0 Condi rion on The index a r rribu re of query mapped ro mul ricas r address Query reaches all objec rs saTisfying condiTion on index a r rribu re Objec rs perform check for remaining condi rions and respond or no r Where To provide nc rionali ry Cl Ne rwor39k versus application layer39 110 Ne rwgrk as Da raSpace engine Space Hande idenTifies daTacube Subjecf Handes are aTTribuTe value pair39s such Tha r a r rr39ibu re is indexed Based on locaTon of dafacube 3058039 on f39eeva f aTTFbufe Da raSpace address 1 11 CMPE 259 Sensor NeTwor39ks Ka ria Obr39aczka Winter 2005 Storage and Querying Overview Overview Cl Sensors sensegenero re do ro CI Usersapplications in reres red in do ro CI Require on infrastructure for do ro access and s roroge Cl Common user opero rions are 0 Queries monitoring 0 Ac ruo re and con rrol Types of queries Cl Hisfor39icali O Wha r is The average rainfall over39 pas r 2 days Cl Current 0 Wha r is The cur39r39en r remper39a re in Rm 226 El Long running 0 Temperature in r39m 226 over39 The nex r 4 hours every 30 seconds Issues D How To iden rify relevon r sensors CI Compu ro rion vs communication rrodeoff 0 Where To process query Inside The sensor neTwork rouTe query AT cenTroIized locoTion rou re doTa Large amounts of data transfer ef ciency 0 How To process query DaTaSDace Queryina and Monitorina Deeply NeTwor39ked CollecTions in Physical SDace T Imielinski and S Goel Rufger39s UniversiTv Cl Billions of objec rs popula re space Cl Each produces and locally s ror39es da ra Cl Loca rion awar39e CI Can be selec rively moni ror39ed quer39ied and controlled 9 Physical wor39ld enhanced wi rh da ra Choroc reris rics Cl Dofospoce O Do ro lives on The objec r 0 Users access no r only local i nformo rion bu r can novigo re en ri re do rospoce O Spo riol world divided in 3D dafacubes CS Bldg s rreeT block eTc O Communico rion messaging and compu ro rion Techniques for querying and moni roring required Qgerying and moni roring Cl Queries are spatially driven Cl S l39eps O Iden rify relevan r da racubes O Iden rify relevan r nodes da raflocks DaTacube direc rory service 0 Aggrega rion for queries on several da racubes Eginforma rion abouT ManhaTTan Taxi cabs Archi rec ring Da raSpace Cl Querying and moni roring O Mul ricas r mechanisms aTTribuTe value pair mapped To a mul ricas r address 0 Group membership based on Physical locaTion A r rribu re Tempera rure vehicles e rc 0 Mapping a r rribu re value To mul ricas r address Eg using hash Tables 0 Condi rion on The index a r rribu re of query mapped ro mul ricas r address Query reaches all objec rs saTisfying condiTion on index a r rribu re Objec rs perform check for remaining condi rions and respond or no r Where To provide nc rionali ry Cl Ne rwor39k ver39sus applica rion layer39 110 Ne rwgrk as Da raSpace engine Space Hande idenTifies daTacube Subjecf Handes ar39e aTTr ibuTe value pair39s such Tha r aTTr ibuTe is indexed Based on locaTon of dafacube 3058039 on f39eeva T aTTFbufe DafaSpace address 1 11


Buy Material

Are you sure you want to buy this material for

25 Karma

Buy Material

BOOM! Enjoy Your Free Notes!

We've added these Notes to your profile, click here to view them now.


You're already Subscribed!

Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'

Why people love StudySoup

Bentley McCaw University of Florida

"I was shooting for a perfect 4.0 GPA this semester. Having StudySoup as a study aid was critical to helping me achieve my goal...and I nailed it!"

Kyle Maynard Purdue

"When you're taking detailed notes and trying to help everyone else out in the class, it really helps you learn and understand the I made $280 on my first study guide!"

Jim McGreen Ohio University

"Knowing I can count on the Elite Notetaker in my class allows me to focus on what the professor is saying instead of just scribbling notes the whole time and falling behind."

Parker Thompson 500 Startups

"It's a great way for students to improve their educational experience and it seemed like a product that everybody wants, so all the people participating are winning."

Become an Elite Notetaker and start selling your notes online!

Refund Policy


All subscriptions to StudySoup are paid in full at the time of subscribing. To change your credit card information or to cancel your subscription, go to "Edit Settings". All credit card information will be available there. If you should decide to cancel your subscription, it will continue to be valid until the next payment period, as all payments for the current period were made in advance. For special circumstances, please email


StudySoup has more than 1 million course-specific study resources to help students study smarter. If you’re having trouble finding what you’re looking for, our customer support team can help you find what you need! Feel free to contact them here:

Recurring Subscriptions: If you have canceled your recurring subscription on the day of renewal and have not downloaded any documents, you may request a refund by submitting an email to

Satisfaction Guarantee: If you’re not satisfied with your subscription, you can contact us for further help. Contact must be made within 3 business days of your subscription purchase and your refund request will be subject for review.

Please Note: Refunds can never be provided more than 30 days after the initial purchase date regardless of your activity on the site.