Directed Diffusion for Wireless Sensor Networking By Chalermek
Directed Diffusion for Wireless Sensor Networking By Chalermek Intanagonwiwat, Ramesh Govindan, Deborah Estrin, John Heidemann, and Fabio Silva Presented by: Jin Sun 02/08/2005 CS 240 Presentation 1
Outline Introduction n The problem n Directed Diffusion Concepts n Simulation Results n Summary n 02/08/2005 CS 240 Presentation 2
Introduction n A region requires eventmonitoring n Deploy sensors forming a distributed network Wireless networking ¨ Energy-limited nodes ¨ n 02/08/2005 CS 240 Presentation On event, sensed and/or processed information delivered to the inquiring destination 3
The Problem n Where should the data be stored? n How should queries be routed to the stored data? n How should queries for sensor networks be expressed? n Where and how should aggregation be performed? 02/08/2005 A sensor field Sensor sources CS 240 Presentation Even t Directed Diffusion Sensor sink On event, sensed and/or processed information delivered to the inquiring destination 4
Directed Diffusion n Initial Goals: ¨ Propose an application-aware paradigm to facilitate efficient aggregation, and delivery of sensed data to inquiring destination 02/08/2005 CS 240 Presentation 5
Directed Diffusion-how it works Low data rate Sink “How many vehicles do you observe in the southeast quadrant? ” n High data rate Source aggregation point Robust, efficient data distribution in sensor networks ¨ ¨ 02/08/2005 Additional source name data (not nodes), use physicality diffuse requests and responses across network optimize path with gradient-based feedback additional data can be processed and aggregated within the network CS 240 Presentation 6
Directed Diffusion Data Naming n Interests and Gradient n Data Propagation n Reinforcement n ¨ Path establishment ¨ Path failure / recovery ¨ Loop elimination 02/08/2005 CS 240 Presentation 7
Data Naming n Expressing an Interest ¨ Using ¨ E. g. , n Data Type = Wheeled vehicle Interval = 20 ms Duration = 10 s Field = [x 1, y 1, x 2, y 2] // detect vehicle location // send events every 20 ms // Send for next 10 s // from sensors in this area reply ¨ Using ¨ E. g. , 02/08/2005 attribute-value pairs Type = Wheeled vehicle // type of vehicle seen Instance = truck // instance of this type Intensity = 0. 6 // signal amplitude measure Confidence = 0. 85 // confidence in the match Timestamp = 01: 20: 34 // event generation time Field = [x 1, y 1, x 2, y 2] // from sensors in this area CS 240 Presentation 8
Directed Diffusion Data Naming n Interests and Gradient n Data Propagation n Reinforcement n ¨ Path establishment ¨ Path failure / recovery ¨ Loop elimination 02/08/2005 CS 240 Presentation 9
Interest Propagation Sink Sources n n n Interest Inquirer (sink) broadcasts exploratory interest, i 1 ¨ Intended to discover routes between source and sink Neighbors update interest-cache and forwards i 1 No way of knowing differentiating new interests from repeated 02/08/2005 CS 240 Presentation 10
Gradient Establishment Sink n Gradient Sink Routed Data Gradient for i 1 set up to upstream neighbor ¨ No source routes ¨ Gradient – a weighted reverse link ¨ Low gradient Few packets per unit time needed 02/08/2005 CS 240 Presentation 11
Directed Diffusion Data Naming n Interests and Gradient n Data Propagation n Reinforcement n ¨ Path establishment ¨ Path failure / recovery ¨ Loop elimination 02/08/2005 CS 240 Presentation 12
Event-data propagation n Event e 1 occurs, matches i 1 in sensor cache ¨ e 1 n identified based on waveform pattern matching Interest reply diffused down gradient (unicast) ¨ Diffusion n initially exploratory (low packet-rate) Cache filters suppress previously seen data ¨ Problem 02/08/2005 of bidirectional gradient avoided CS 240 Presentation 13
Directed Diffusion Data Naming n Interests and Gradient n Data Propagation n Reinforcement n ¨ Path establishment ¨ Path failure / recovery ¨ Loop elimination 02/08/2005 CS 240 Presentation 14
Reinforcement Reinforced gradient D Event Reinforced gradient B A sensor field Sink A C n From exploratory gradients, reinforce optimal path for high-rate data download Unicast ¨ By requesting higher-rate-i 1 on the optimal path ¨ Exploratory 02/08/2005 gradients still exist – useful for faults CS 240 Presentation 15
Path Failure / Recovery n Link failure detected by reduced rate, data loss ¨ Choose next best link (i. e. , compare links based on infrequent exploratory downloads) n Negatively reinforce lossy link ¨ Either send i 1 with base (exploratory) data rate ¨ Or, allow neighbor’s cache to expire over time D Event M Src C 02/08/2005 A B CS 240 Presentation Sink Link A-M lossy A reinforces B B reinforces C … D need not A negative reinforces M M negative reinforces D 16
Loop Elimination P D n Q M A M gets same data from both D and P, but P always delivers late due to looping ¨M negatively-reinforces (nr) P, P nr Q, Q nr M ¨ Loop {M Q P} eliminated n Conservative nr useful for fault resilience 02/08/2005 CS 240 Presentation 17
Simulation Results n Compare directed diffusion to ¨ flooding ¨ Omniscient n multicast Key metrics: ¨ Average dissipated energy per node energy dissipation / # events seen by sinks ¨ Average packet delay latency of event transmission to reception at sink ¨ Distinct event delivery # of distinct events received / # of events originally sent 02/08/2005 CS 240 Presentation 18
Average Dissipated Energy flooding Multicast Diffusion In-network aggragation reduces DD redundancy - Flooding is poor because of multiple paths from source to sink 02/08/2005 CS 240 Presentation 19
Delay flooding Diffusion Multicast DD finds least delay paths - Floof]ding incurs latency due to high MAC contention, colission 02/08/2005 CS 240 Presentation 20
Event Delivery Ratio under node failures 0% 20% 10% Delivery ration degrades with more nodes failures - Graceful degradation indicate efficient negative reinforcement 02/08/2005 CS 240 Presentation 21
Summary n Main Contributions ¨ Description of new networking paradigm n Interests, gradients, reinforcement n Benefits of in-network processing n Aggregation and nested-queries ¨ Works with multiple sources and sinks ¨ Can perform local repair ¨ Reinforce another path if a node dies 02/08/2005 CS 240 Presentation 22
Summary (cont’d) n Disadvantages ¨ Design doesn’t deal with congestion or loss ¨ Periodic broadcasts of interest reduces network lifetime ¨ Nodes within range of human operator may die quickly 02/08/2005 CS 240 Presentation 23
Thank You! 02/08/2005 CS 240 Presentation 24
- Slides: 24