Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks - PowerPoint PPT Presentation

event query processing based on data centric storage in wireless sensor networks n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks PowerPoint Presentation
Download Presentation
Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks

play fullscreen
1 / 33
Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks
152 Views
Download Presentation
fox
Download Presentation

Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Event Query Processing Based on Data-Centric Storage in Wireless Sensor Networks Longjiang Guo, Yingshu Li and Jianzhong Li Globalcom IEEE 2007

  2. Outline • Introduction • Event formulation and storage • Event query processing • Choosing optimal storage strategy • Simulation • Conclusion

  3. INTRODUCTION • interested in the events, instead of the sensors • A event is a fusion of multiple sensed data satisfying some conditions. • event fire: temperature>90 AND smoke>100L/mol, • find all the locations where ET satisfied the pre-defined observation attributes.

  4. EVENT FORMULATION AND STORAGE • preprocessing of the observed data : • includes a check on whether the data satisfies the event conditions • If it does, the event information will be stored in the sensor network.

  5. EVENT FORMULATION AND STORAGEA. Event Formulation • Definition 1: Observation attribute Ai (i=1…n) • Definition 2: Observation value • Definition 3: Round • Definition 4: Event type ETi (A1, A2, …, Aiki ) • a group of pre-defined observation attributes • ET(Tmpt, Wnd, Hudit) = Tmpt>t1 and Wnd>w2 and Hudit>h3 • Definition 5: Observation zone • divided into m×n grids ,each grid is an observation zone. • Definition 6: Event e(ET, g, t) • Definition 7: Event query Q{[t1, t2],ETi}

  6. EVENT FORMULATION AND STORAGEA. Event Formulation • The observation nodes ofETi: Si1, Si2, …, Siki , wherenode Sijobserves attribute Aij. • W: types of eventsET1 ,ET2 ,…,ETw

  7. EVENT FORMULATION AND STORAGEA. Event Formulation • Every node Sijbroadcasts a message within its observation zone. • ETi , Sij’s ID, Sij’sposition • When a node Sijreceives a message, • it firstly checks whether the event type is the same as its own • forwards the message to its neighbors

  8. EVENT FORMULATION AND STORAGEA. Event Formulation • For each observation zone, in the kth round, Sijis selected as the event fusion node • j=(kmod ki )+1 the event fusion node routes the event to storage node k=1, Si1 as event fusion node Si2 ,…, Si6 route IDs and values to Si1 Si2 Si6 Si1 Broadcast Si1 ‘s position Si5 Si3 Si4

  9. EVENT FORMULATION AND STORAGEB. Event Storage • External storage: • the sink is the event storage node • an event fusion node will route the events directly to the sink • Local storage: • the event storage node is the same as the event fusion node. • Data centric storage: • hash to an in-network position (x, y) according to the event type. • nearest to that position, called Home Node

  10. EVENT FORMULATION AND STORAGEB. Event Storage • More energy can be saved to put Home Nodeat the center of a network. • Center Mapping Data Centric Storage (CM-DCS) • Events of the same type will be hashed to an observation zone lying at the center of the network. • A node in the observation zone nearest to the center of the network serves as an event storage node

  11. EVENT FORMULATION AND STORAGEB. Event Storage • Lemma 1: • the total energy consumption for routing events is proportional to the distance between the event fusion node and the event storage node. • Proof: This is obvious.

  12. EVENT FORMULATION AND STORAGEB. Event Storage • Lemma 2: • If observation nodes are distributed uniformly in an observation zone, • event fusion node is the center of the observation zone. • Proof: event fusion node’s position (Xf, Yf) : E(Xf)=0.5(a+c) E(Yf)=0.5(b+d) an observation node(X, Y) is a two dimensional random variable. (c, d) (E(Xf), E(Yf)) (a, b)

  13. EVENT FORMULATION AND STORAGEB. Event Storage • Theorem 1: • If event storage nodes are located near the center of the sensor network, • energy consumption for routing events from an event fusion node to an event storage node is minimized. • Proof: • (X1, Y1), (X2, Y2), …, (Xmxn, Ymxn) :centers of the observation zones (fusion nodes). • (X, Y) : event storage node 

  14. EVENT FORMULATION AND STORAGEB. Event Storage Mapping: {ET1→Z6; ET2→Z10;ET3→Z11;ET4→Z7} : storage node

  15. EVENT FORMULATION AND STORAGEC. Storage at an Individual Event Storage Node • each event storage node is responsible for event type ETi • time-stamped vector-based storage strategy: • If (j1, j2, …, jk of I)= 1 , • observation zones detected the event of type ETi.

  16. EVENT QUERY PROCESSING • extract information from the sensor network for query Q{[t1, t2], ETi}. • Event query processing based on CM-DCS • Event query processing based on Local Storage

  17. EVENT QUERY PROCESSINGA. Event query processing based on CM-DCS • Phase 1: Deciding the routing destination. • Phase 2: Routing query Q{[t1, t2], ETi}. from the sink to the event storage node p • Phase 3: Answering query Q{[t1, t2], ETi}. • Phase 4:Routing Ipback to the sink from node p. A={(I,t)} | t [t1 , t2 ], then sink Q{[t1 , t2], ETi}. Ip (xi , yi) p ETi

  18. EVENT QUERY PROCESSINGB. Event query processing based on Local Storage • Phase 1:Query dissemination. • sink broadcasts the query, add its ID to query packet. • p replaces the ID in the query packet with its own ID • Phase 2:Collection of children’s IDs. • p broadcasts <p,fp> , where fp is p’s parents • p receives {<q1, fq1>, < q2, fq2>, …, < qm, fqm>}, Children(p)={qi| <qi, fqi>, where fqi=p}. • Phase 3:Combination of query results. • event storage node preceives an event query • A={(I,t)} | t [t1 , t2 ], then pcomputes • Phase 4:Routing Ipback to the sink from node p.

  19. CHOOSING OPTIMAL STORAGE STRATEGY • provide a guideline of choosing a correct storage strategy for different applications.

  20. CHOOSING OPTIMAL STORAGE STRATEGY

  21. CHOOSING OPTIMAL STORAGE STRATEGY • A. Estimation of Energy Consumption For Initialization • broadcast a message and forward a message to its neighbors Receive from neighbors/broadcast to neighbors

  22. CHOOSING OPTIMAL STORAGE STRATEGY • B. Estimation of Energy Consumption For Event Formulation • The expected number of hops (nodefusion node) =sqrt(N/(8mn)) (A random node fusion node)

  23. CHOOSING OPTIMAL STORAGE STRATEGY • C. Estimation of Energy Consumption For External Storage • The expected number of hops (nodesink) =(2N)/2 #events (fusion node sink) (node fusion node)

  24. CHOOSING OPTIMAL STORAGE STRATEGY • D. Estimation of Energy Consumption For CM-DCS • The expected number of hops (node storage node) =(2N)/4 (node fusion node) Query: (sink storage node) Answer: (storage node sink) (fusion node storage )

  25. CHOOSING OPTIMAL STORAGE STRATEGY • E. Estimation of Energy Consumption For Local Storage Query Answer

  26. CHOOSING OPTIMAL STORAGE STRATEGY • F. Comparing the Energy Consumptions • Assume: Eb=Eu, Er =1.5Eb , Sevent =Squery =0.25Sresult • Observation 1: if Nq>N·prob, external storage • Observation 2: • ifρ and Nqincrease,CM-DCS • if ρ is a constant and Nincreases, local storage

  27. SIMULATION RESULTS

  28. SIMULATION RESULTS

  29. SIMULATION RESULTS

  30. SIMULATION RESULTS

  31. SIMULATION RESULTS

  32. SIMULATION RESULTS

  33. CONCLUSION • propose a data centric storage strategy CM-DCS • eventquery processing algorithms: EP-CM-DCS and EP-LS. • compare the energy consumptions • users can have a guideline of choosing a correct storage strategy for different applications.