Structurefree Data Aggregation Kaiwei Fan Sha Liu and
- Slides: 22
Structure-free Data Aggregation Kaiwei Fan, Sha Liu, and Prasun Sinha (speaker) The Ohio State University Dept of Computer Science and Engineering
Outline n n n Introduction Structure-free Data Aggregation Simulation Results Experiments on a testbed Conclusion
Introduction n Data Aggregation q q n In-network processing Reduces communication cost Approaches q Static Structure n n q [LEACH, TWC ’ 02] [PEGASIS, TPDS ’ 02] Dynamic Structure n n [Directed Diffusion, Mobicom ‘ 00] [DCTC, Infocom ‘ 04]
Static Structure n Pros q q n Low maintenance cost Good for unchanging traffic pattern Cons q Unsuitable for event triggered network n n Long link-stretch Long delay sink
Static Structure n Pros q q n Low maintenance cost Good for unchanging traffic pattern Cons q Unsuitable for event triggered network n n Long link-stretch Long delay sink
Dynamic Structure n Pros q n Reduces communication cost Cons q High maintenance overhead sink
Structure-free Data Aggregation n Challenge q q n Approach q q n Routing: who is the next hop? Waiting: who should wait for whom? Spatial Convergence Temporal Convergence Routing? Waiting? Solution q q Data Aware Anycast Randomized Delay sink
Data Aware Anycast n n Improve Spatial Convergence Anycast q n One-to-Any forwarding scheme Anycast for Immediate Aggregation q q To neighbor nodes having packets for aggregation Keep Anycasting for Immediate Aggregation sink
Data Aware Anycast 50 nodes in 200 mx 200 m sink
Data Aware Anycast n Forward to Sink q q To neighbor nodes closer to the sink Using Anycast for possible Immediate Aggregation sink
Data Aware Anycast n Forwarding and CTS replying priority q q q Class A: Nodes for Immediate Aggregation Class B: Nodes closer to the sink Class C: Otherwise, do not reply mini-slot Sender Class A Nbr Class B Nbr Class C Nbr RTS CTS slot Class A Class B CTS Canceled CTS
Randomized Waiting n n Improve Temporal Convergence Naive Waiting Approach q q n Use delay based on proximity to sink (closer to sink => higher delay) Long delay for nodes close to the sink in case the event is near the sink Our Approach: Random Delay at Sources
Analysis n n Sink Y: Number of hops a packet is forwarded before being aggregated Assumptions: q q Each node has k choices for next hops closer to sink All n nodes have packets to send …… h=n/k n E[Y] = q x : random delay in [0, 1] picked up by a node q dh : random delay chosen by a node h hops away from sink n Total Number of Transmissions =
Analysis vs. Simulation n Results matches up to 40 hops Gap increases as network size increases Reason: transmission delay is ignored in analysis
Simulation Results n Evaluated Protocols q q q Opportunistic (OP) Optimum Aggregation Tree (AT) Data Aware Anycast (DAA) Randomized Waiting (RW) DAA+RW n Evaluated Metric q n Normalized Number of Transmissions Parameters Studied q q Maximum Delay Event Size Aggregation Function Network Size
Simulation Results – Maximum delay n Configuration q q q 33 x 33 grid network event moves at 10 m/s event radius: 200 m 140 nodes triggered by the event data rate: 0. 2 pkt/s data payload: 50 bytes n AT-2: Aggregation tree approach with varying delay n DAA+RW improve OP by 70%
Simulation Results – Maximum delay n n AT is sensitive to delay AT has best performance with highest delay
Simulation Results – Event Size n Configuration q q q n event radius: 50 m ~ 300 m 8 ~ 260 nodes triggered by the event radius: 200 m Key Observations q DAA+RW is much better than OP q DAA+RW is close to AT (optimal tree)
Simulation Results – Aggregation Ratio n Configuration q q q n Aggregation Ratio ρ: 0~1 Packet size: max(50, 50* (1 -ρ)* n) Max packet size: 400 bytes Key Observation q DAA+RW performs better than AT q Following the best tree is not optimum if the packet size is limited
Simulation Results – Network Size n event distance to the sink: 300 m ~ 700 m event radius: 200 m n Key Observation n q Improvement is higher for events farther from the sink
Experiment – Randomized Waiting n Linear network with 5 sources and 1 sink 0. 2 pkt/s data payload: 29 bytes n Key Observation n n q Delay as low as 0. 1 is sufficient for optimizing performance
Conclusion n Data Aware Anycast for Spatial Convergence Randomized Waiting for Temporal Convergence Efficient Aggregation without a Structure q q High Aggregation No maintenance overhead
- What is fanout in vlsi
- Alex liu cecilia liu
- Líu líu lo lo ta ca hát say sưa
- Deu nexia fan sensornexia fan sensor
- Kai-wei chang
- Kaiwei chang
- Kaiwei chang
- Aggregator service in microservices
- Ramackas
- What is logistics network configuration
- Data pre processing
- Composition uml
- Difference between aggregation and composition
- Pods aggregation and silos in cloud computing
- Aggregation in system analysis and design
- Computing refers to
- Cho sha cao
- Domain of tangent function
- Staying healthy assessment
- Sha questionnaire
- Sha 128
- Master sha
- Ntu