Flash Player 9 (or above) is needed to view presentations.
We have detected that you do not have it on your computer. To install it, go here.

Like this presentation? Why not share!

Ad Hoc Now2008 Probabilistic Query Dissemination



Talk presented at Ad Hoc Now 2008 in Nice.

Talk presented at Ad Hoc Now 2008 in Nice.



Total Views
Views on SlideShare
Embed Views



1 Embed 1

https://www.linkedin.com 1



Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

Ad Hoc Now2008 Probabilistic Query Dissemination Ad Hoc Now2008 Probabilistic Query Dissemination Presentation Transcript

  • Zinaida Benenson Felix Freiling Markus Bestehorn Marek Jawurek Erik Buchmann Query Dissemination with Predictable Reachability and Energy Usage in Sensor Networks AdHoc-Now 2008, Sophia Antipolis www.kit.edu
  • Introduction – Sensor Networks Introduction A sensor network consists multiple of sensor nodes, e.g. Problem Desc. Idea Id Reachability Direct Indirect I di t Topology MicaZ Sun SPOT Information Evaluation Sensor Nodes Setup Simulation Battery-powered Break Even Break-Even Equipped with sensor hardware Deployment Limited computing resources Conclusion Wireless communication Q&A Markus Bestehorn Slide 2
  • Query Processing in WSN Introduction Generic query processing in sensor networks approach: Problem Desc. 1. Disseminate query through base station q y g Idea Id SELECT MAX(temp) FROM sensors … Reachability 2. Measure data using sensing hardware Direct Indirect I di t 3 Process & route query results back to base station 3. Topology Information Optimization Goal: Reduce energy consumption! Evaluation Sending/Receiving data most expensive! Setup Simulation 15°C Break Even Break-Even 2 17°C 21°C Deployment 4 Q 6 Conclusion Q Q Q Q&A 22°C 3 19°C Q 5 Basestation 22°C 1 20°C Markus Bestehorn Slide 3
  • Challenges for Query Dissemination Introduction Unnecessary rebroadcasts must be avoided Problem Desc. Nodes should receive query only once q y y Idea Id Q Q Reachability 2 4 6 Direct ? Indirect I di t 1 3 Q Topology 5 Q Information Existing approaches Evaluation Setup Topology-Based: Determine rebroadcasting nodes using Simulation accurate local topology information Break Even Break-Even 2-Hop topology information is very costly Deployment Optimal Broadcast Dominating Set Problem NP-complete Conclusion Probabilistic: Nodes rebroadcast with probability p Q&A High p high energy consumption Low p not all nodes reached How to set p? Markus Bestehorn Slide 4
  • Idea & Agenda Introduction General idea: Problem Desc. Acquire basic topology information q p gy Idea Id does not consume as much energy Reachability Use probabilistic approach to disseminate query Direct Indirect I di t Set rebroadcast probability based on basic topology information Topology Information Agenda: Evaluation Prediction framework Setup How to predict reachability for a given rebroadcast probability p? Simulation Break Even Break-Even How to set p based on prediction to reach all nodes? Deployment Topology Discovery Conclusion Possibilities to aquire required topology information? Extensive evaluation Q&A Simulation and real deployment results Explore tradeoff reachability vs. energy consumption Accuracy of the Prediction F A f th P di ti Framework? k? Markus Bestehorn Slide 5
  • Hop Set Modell (1) Introduction Task: Predict the number of reached nodes given Problem Desc. Topology information p gy Idea Id Rebroadcast probability p Reachability Direct Hop Set: Hop Set H[i] contains all nodes that can be Indirect I di t reached by the base station via i hops Topology H[3] H[2] H[1] H[0] Information Evaluation Setup 2 4 6 Simulation Break-Even Break Even Deployment 1 3 5 Conclusion Q&A Markus Bestehorn Slide 6
  • Hop Set Modell (2) Introduction Possibilities to reach a node via broadcast Problem Desc. Direct: Message is sent from node in H[i-1] to node in H[i] g [ ] [] Idea Id Indirect: Message is sent from node in H[i] to node in H[i] Reachability Direct Backwards: Node in H[j] with j > i forwards message to node in Indirect I di t H[i] Simplification: not considered Topology Information Evaluation Setup H[3] H[2] H[1] H[0] Simulation Break-Even Break Even Q 4 Deployment 2 6 Conclusion Q Q Q&A 1 3 5 Markus Bestehorn Slide 7
  • Reachability Prediction Introduction R(h,p) := number of reached nodes in Hop Set h with Problem Desc. rebroadcast probability p p y Idea Id R(0,p) = 1 base station is always „reached“ Reachability Direct R(1,p) = |H[1]| Indirect I di t base station always broadcasts H[1] Hop Set H[1] always reached Topology Information Evaluation Nodes in s bseq ent Hop Sets are reached subsequent Setup Directly Direct(h,p) Simulation Example: Direct(2,p)=4 Break Even Break-Even Indirectly Indirect(h p) Indirect(h,p) H[1] Deployment Example: Indirect(2,p)=2 H[2] Conclusion Q&A R(h,p) := Direct(h,p) + Indirect(h,p) with h > 1 Markus Bestehorn Slide 8
  • Direct Reachability Prediction Introduction Basic Idea to compute Direct(h,p) H[i-1] H[i] Problem Desc. Possible rebroadcasters | [ ]| |H[h-1]| nodes Idea Id Potential Rebroadcasters R(h-1,p) nodes Reachability Direct Rebroadcasters R(h-1,p)·p nodes Indirect I di t |H[h-1]| |H[h 1]| R(h-1,p) R(h 1 ) Topology R(h-1,p) ·p Information Evaluation Setup Simulation P(„Node in H[h] directly reached“) can be computed Break Even Break-Even Deployment Avg. Number of connections from H[i] to H[i-1] Connectivity[h] Conclusion Detailed description in the p p p paper Rebroadcast Q&A p Probability Direct(h,p) = P(reached directly)·s[h] Nodes reached H[i] in i Hops p Markus Bestehorn Slide 9
  • Indirect Rechability Prediction Introduction Idea to compute Indirect(h,p): Problem Desc. Potential Rebroadcasters Direct(h,p) ( ,p) Idea Id Rebroadcasters Direct(h,p)·p Reachability Direct Average Number of connections within a Hop set Indirect I di t Interconnectivity[h] Topology Indirect(h,p)=Direct(h,p)·p·Interconnectivity[h] Information H[2] H[1] H[0] Evaluation Setup 4 6 Simulation Break Even Break-Even Deployment 3 5 Conclusion Implicit Assumption: Rebroadcast Q&A p Reached nodes distributed Probability evenly within hop sets H[i] Nodes reached in i Hops p Markus Bestehorn Slide 10
  • Reachability Prediction (3) Introduction R(h,p) computes reached nodes in Hop Set h with Problem Desc. rebroadcast probability p p y Idea Id Computing total reachability for given p: Reachability Direct Indirect I di t R ( p ) = ∑ min (R(h, p ), H [h]) Topology h Information Minimum required because Direct(h p) + Indirect(h p) > H[h] Direct(h,p) Indirect(h,p) Evaluation possible Setup Simulation Break Even Break-Even Deployment Also available: Conclusion Number of sent messages / rebroadcasting nodes Q&A Number of received messages Allows estimation of energy consumption! Markus Bestehorn Slide 11
  • Topology Information Introduction Required Topology Information for Reachability Prediction Problem Desc. Set Size: Number of Nodes in each Hop Set H[h] p [ ] Idea Id Connectivity: Avg. Number of connections a node in H[h] has Reachability to nodes in H[h-1] Direct Indirect I di t Interconnectivity: Avg. Number of connections a node in H[h] Avg has to other nodes in H[h] Topology Information Example: Evaluation Setup Simulation H[i-1] H[i] Set size Connectivity Break-Even Break Even Deployment … i-1 i … … i-1 i … … 2 3 … … 1.5 2 … Conclusion Q&A Interconnectivity I t ti it … i-1 i … … 0 4/3 … Markus Bestehorn Slide 12
  • Acquiring Topology Information Introduction Several options to get required topology information: Problem Desc. Echo Algorithm g Idea Id Expansion Wave: Explore network by initiating a flooding at the Reachability base station Direct Contraction Wave: Aggregate topology information towards base gg g p gy Indirect I di t station Topology Drawback: Energy consumption, Scalability Information Gossiping: Nodes attach routing information to messages Evaluation Advantage: No extra messages Setup Drawback: Routing information disperses slowly Simulation Break Even Break-Even Routing Protocol Extraction: Extract topology information Deployment from data structures of routing protocol Conclusion Drawback: Only possible for some protocols (AODV) Q&A Note: N t Even for Echo Algorithm Prediction pays off after a few q y query disseminations! Markus Bestehorn Slide 13
  • Evaluation - Setup Introduction Network: 125 to 425 nodes Problem Desc. Node Degree: 4 – 16 g Idea Id Different Topology Types used, e.g. Reachability Direct Uniform: Nodes are placed uniformly around basestation Indirect I di t Gaussian: G G i Gaussian di ib i of nodes around b i distribution f d d basestation i Topology 100 topologies per topology type, 40 queries per topology Information Energy prediction based values measured on MicaZ Evaluation Setup Simulation Criteria for success: Break Even Break-Even Deployment Accurate Prediction for Reachability and Energy Optimization of probabilistic rebroadcast parameter p Conclusion to reach ALL nodes with query Q&A without rebroadcasting at each node Exploration of rebroadcast probability – reachability tradeoff Markus Bestehorn Slide 14
  • Evaluation – Simulation Results Introduction Result for node degree 16, 425 nodes Problem Desc. Idea Id Uniform Gaussian Reachability Direct Indirect I di t Topology Information p0 Evaluation Setup Simulation Break Even Break-Even Deployment Conclusion Findings: Q&A Reachability & energy prediction accurate For most experiments, there exists a p0<1: Increasing p beyond p0 does not pay off regarding reachability! p y g g y energy savings without reducing reachability Markus Bestehorn Slide 15
  • Break Even Point Introduction Exemplary computation: Problem Desc. Uniform topology p gy Idea Id 425 nodes, node degree 16 Reachability Direct Assuming Indirect I di t Topology di T l discovery using the E h Al i h i h Echo Algorithm Topology Energy consumption values measured on MicaZ Information Evaluation Setup Topology Discovery consumes 722 mAs Simulation Query dissemination with simple flooding (p=1) consumes Break Even Break-Even 370 mAs A Deployment Using prediction framework for 99% reachability Conclusion p p=0.6 220 mAs Q&A Result: Topology Discovery pays off after 5 queries! Markus Bestehorn Slide 16
  • Evaluation – SPOT Deployment Introduction 17 SPOTs + Basestation deployed Problem Desc. Idea Id 10 Queries were disseminated into the network using Reachability Direct Simple flooding (p=1) Indirect I di t Probabilistic flooding P b bili ti fl di Topology Prediction algorithm was used to reach Information All nodes Evaluation At lowest possible rebroadcast prob. p Setup Simulation Result: Break Even Break-Even Broadcast Reached Sent Msg Msg. Received Deployment Algorithm Nodes Msg. Simple 16.3 16.3 63.8 Conclusion Probabilistic 15.4 15 4 10.2 10 2 34 Q&A Probabilistic Rebroadcast Optimization ~30% less sent messages almost 50% less received messages Markus Bestehorn Slide 17
  • Summary Introduction Explored relations between Problem Desc. Reachability y Idea Id Energy consumption for query dissemination Reachability Direct Energy spent to acquire topology information Indirect I di t Introduced analytical f I t d d l ti l framework k Topology Determines p0<1 for probabilistic broadcasting to reach all Information nodes Evaluation Allows predictions regarding Setup sent / received messages Simulation Energy consumption gy p Break Even Break-Even Deployment Energy spent for topology information pays off after a few (5) Conclusion query disseminations Even if echo algorithm is used! Q&A Evaluation using Simulation & real Sensor network Markus Bestehorn Slide 18
  • Outlook Introduction Integrate „backwards“ reachability into Problem Desc. framework Idea Id More topology information required? Reachability Direct Payoff? Indirect I di t Relation between query dissemination and query result Topology accuracy Information Evaluation <100% reachability ~100% reachability ? accuracy 100% accuracy Setup p0 Simulation Break Even Break-Even Deployment Dynamic usage of different broadcast algorithms Conclusion Probabilistic approach good for dense networks Q&A Switch to other broadcast algorithms in less populated areas of the network? Markus Bestehorn Slide 19
  • Thank you for your attention! Introduction Problem Desc. Idea Id Reachability Direct Questions? Indirect I di t Topology Information Evaluation Setup Simulation Break-Even Break Even Deployment Conclusion Q&A Markus Bestehorn Slide 20