/42 Does Wireless Sensor Network Scale? A Measure Study on GreenOrbs Yunhao Liu, Yuan He, Mo Li, Jiliang Wang,Kebin Liu, Lufeng Mo, Wei Dong, Zheng Yang, Min Xi, Jizhong Zhao, Xiang-Yang Li Presented by: 张 猜
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 What's wireless sensor networks? Components: A group of specialized transducers A communications infrastructure
/42 Applications Environment monitoring Scientific observation Emergency detection Field surveillance Structure monitoring......
/42 Existing Large-scale sensor networks VigilNet for military surveillance fig.1 The VigilNet system Architecture
/42 Existing Large-scale sensor networks Motelab for indoor testbed fig.2 Component model of Motelab
/42 Existing Large-scale sensor networks SensorScope for weather monitoring fig.3 Design of a sensing station.
/42 Existing Large-scale sensor networks Trio based on solar-powered fig.4 The system hardware architecture consists of four tiers: mote, gateway, server, and client.
/42 Existing Large-scale sensor networks Drawbacks of these deployments Often highly optimized for Specific application needs Not fully leveraged as Platforms for consistently observing general network behaviors.
/42 An overlook on the sensor network deployment and the real topology
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 Tens of sensors A Macroscope in the Redwoods Fidelity and yield in a volcano monitoring sensor network SensorScope: Out-of-the-Box Environmental Monitoring
/42 Tens of sensors Drawbacks: Hardly reveal some network behaviors Routing dynamics and Topology evolution
/42 Sensor Networks at scale ExScal: elements of an extreme scale wireless sensor network Trio: Enabling Sustainable and Scalable Outdoor Wireless Sensor Network Deployments
/42 Sensor Networks at scale Not proper for Measure study Organized hierarchically Either Large-scale or long-lifetime,but not both
/42 Measurements of GreenObrs! Fairly Comprehensive Low-level –Radio signal Strength –Link quality High-level –Routing –Data traffic
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 GreenOrbs Overview System and Applications –Canopy closure estimate –Research on biodiversity –Carbon sequestration –Fire risk prediction
/42 GreenOrbs Overview
/42 TinyOS 2.1
/42 GreenOrbs Overview Data Set –In December 2009 –29 consecutive days – data packets
/42 Data Set GreenOrbs Overview
/42 GreenOrbs Overview Data Set –Back End Data Set Refers to the entire data set collected at the sink –Routing trace –Link trace –Node statistics trace –Overhearing –Beaconing –Local Logging
/42 Measure and Derivations –Yield –Packet Reception Ratio(PRR) / Loss Ratio(PRL) –Packet Delivery Ratio (PDR)
/42 GreenOrbs Overview Measure and Derivation –End-to-end delay –Correlation coefficient A statistical measure of association between two varialbe, e.g. ETX and PDR. Range: [-1,1]
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 Basic Observation Network Characteristics
/42 Basic Observation Network characteristics
/42 Basic Observation Network Characteristics To investigate cause of packets losses: –Three categories of packet losses Transmit_Timeout (61.08%) Receive_Pool_Overflow (38.92%) Send_Queue_Overflow (0%)
/42 Basic Observation Network Characteristics
/42 Basic Observation Investigating Critical Nodes/Links
/42 Basic Observation Investigating Critical Nodes/Links
/42 Basic Observation Looking into the Links
/42 Basic Observation Looking into the Links
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 Who moved Our Cheese The Last Straw that Breaks the Camel’s Back First Conjecture: The bottlenecks are intermediate nodes instead of those lying in the “hot area” around the sink.
/42 Who moved Our Cheese How Dynamic Is the Environment? Second Conjecture: Most of the wireless links used in sensor networks are physically stable.The dynamics mainly come from internal network operation.
/42 Who moved Our Cheese Adaptive Routing Design Third Conjecture: The environment: unpredictable impact on the sensor network system running under it. Current dynamic routing approach: Lack adaption to environment.
/42 Outline Introduction Related Work GreenOrbs Overview Basic Observation Who moved Our Cheese Conclusion
/42 Conclusion Conduct a long term and large-scale measurement study on an operating sensor network in the wild. Validate three conjectures for future sensor network design
/42 Thank you! Q & A