Data Dissemination in Sensor Networks Challenges and Solutions by Sovrin Tolia.
Transcript of Data Dissemination in Sensor Networks Challenges and Solutions by Sovrin Tolia.
Data Dissemination in Sensor Networks
Challenges and Solutionsby
Sovrin Tolia
Presentation Outline Overview of Sensor Networks Directed Diffusion, A Scalable and
Robust Communication Paradigm for Sensor Networks
SPIN, Adaptive Protocols for Information Dissemination in Wireless Sensor Networks
Research Efforts Directed Diffusion
USC/ISI Institute Deborah Estrin, Ramesh Govindan
et al SPIN Protocols
MIT Balakrishnan, Joanna Kulik et al
Challenges in Sensor Networks
Sheer Numbers Unattended deployment Device failures Frequent Change Heterogenity of Tasks
Sensor Nodes Capability of Sensor Nodes Distributed Microsensing Sensor Networks are task specific Autonomous Operation
Sensor Network Applications
Monitoring Remote Geographic Locations and terrains
Toxic Urban Locations Less Accessible environments like
large Industrial Plants and Aircraft Interiors
Directed Diffusion Motivated by Scalability, Energy
Efficiency and and robustness Data Centric Application Aware Energy Efficient Remote Surveillance Sensor
Network
Directed Diffusion Naming
Directed Diffusion Interests and Gradients
Sink Nodes Periodic Broadcasts Interest Cache in every node (distinct
interests)Two Way Gradient establishedGeneralized Interest Propagation means
Directed Diffusion Data Propagation location satisfied, finds Interest
Match from cache computes the highest data rate in
gradients present in the interest cache
Intermediate Node base their decision based on Data cache
“Downconversion”
Directed Diffusion Reinforcement
Data Driven Rules for deciding which node(s) to reinforce
Explicitly send a high data rate Interest message
What do the neighboring nodes do Extreme Reactivity and Negative
Reinforcements
Directed Diffusion Explained
In Essence No mention of target detection algos The usage is not restricted to
interests-gradients-drawing data. All Communication uses interests to
specify named data Neighbor-Neighbor (Not End-to-End) No routers, governed by task
specificity No Need for globally unique identifiers
Is It Routing ? Kind of Reactive, Routes on Demand No Attempt to find one loop free
route Constrained Flooding is used Message Cache is used to avoid loops
They did carry out some preliminary results
Future Research here Target Detection Algorithm Simulate Congestion in the network Experiment the non-aggressive
reinforcement strategy and less conservative negative reinforcements
Sensor Networks are at a stage where Internet was three decades ago
SPIN, Sensor Protocols for Information via Negotiation Overcomes Implosion, Overlap and
Resource Blindness in conventional approaches
Advocates Negotiation and Resource-adaptation
SPIN Naming Using Meta Data SPIN Messages
ADV,REQ,DATA Resource Management Implementation 3-stage handshake Participation only when it can complete
all the three stages
SPIN Working
SPIN Architecture
Performance Compared against flooding, gossiping
and ideal protocol Metric is amount of data disseminated
over time and energy consumed SPIN-1 higher throughput than gossip
but same as flooding, less energy usage
SPIN-2 ideal amount of data per unit energy
Food for Thought If Ad-hoc networks are not yet a
reality, can we ever expect Sensor Networks to be deployed
Are they really required ?
ANY QUESTIONS