1. 1 CS 268: Lecture 24 Sensor Network Architecture (SNA) Ion Stoica Computer Science Division Department of Electrical Engineering and Computer Sciences University of California, Berkeley Berkeley, CA 94720-1776

2. 2 Sensor Network Protocols Today Phy Link Topology Routing Transport Appln RadioMetrix RFM CC1000 Bluetooth 802.15.4 eyes nordic WooMac SMAC TMAC WiseMAC FPS MintRoute ReORg PAMAS CGSR DBF MMRP TBRPF BMAC DSDV ARADSR TORA GSR GPSR GRAD Ascent SPIN SPAN Arrive AODV GAF Scheduling Resynch Yao Diffusion Deluge Trickle Drip Regions Hood EnviroTrack TinyDB PC TTDD Pico FTSP Obligatory David Culler Slide… What if I want to use any two protocols together??

3. 3 Network Model Transit Network Basestation Sensor Patch Patch Network Base-Remote Link Data Service Internet Client Data Browsing and Processing Sensor Node Gateway  Dense patches of sensing nodes -Many resource constrained -Non-homogeneous -Modalities, roles, HW, SW -Power, BW  Transit tier -Often specialized wireless -Provides gateways  Internet Tier -Multiple connections to infra -Deep storage, proc. Viz  SNA should not require unconstrained nodes  Should utilize unconstrained nodes to reduce burden on constrained ones  Mobility within physically embedded context

4. 4 What is an Architecture?  Architecture is how to “organize” implementations -What interfaces are supported -Where functionality is implemented  Architecture is the modular design of the network  Architecture is not the implementation itself

5. 5 Internet vs Sensor Nets Internet goals  Interconnect separate networks  Resilience to loss and failure  Support many comm. services  Accommodate variety  Distributed management  Cost effective  Low effort attachment  Resource accountability  Network Architecture Sensor Nets  Resource efficiency  Data centric design  Deal with intermittent connectivity  Self-managed  Observation, monitoring of various environments  Cost effective  Scalability

6. 6 Internet vs Sensor Nets Internet goals  Interconnect separate networks  Resilience to loss and failure  Support many comm. services  Accommodate variety  Distributed management  Cost effective  Low effort attachment  Resource accountability  Network Architecture Sensor Nets  Dense real world monitoring  Resilience to loss, failure and noise  Support many applications  Scale to large, small, long  Cost effective  Evolvable in resources  Composable  Security

7. 7 Why not IP?  One or very few applications running on a sensornet vs huge number running in the Internet  Large variety of traffic patterns (most not point-to-point): -Any-to-any, many-to-one, many-to-few, one-to-many -Inneficient to impl. these patterns over point-to-point  IP does not address (well): -Resource and energy constraints -Unattended operation -Intermittent connectivity -Space embeded nodes -...

8. 8 A Sensor Network Architecture (SNA)  Narrow waist: Sensornets Protocol (SP) -Goals: generality and efficiency -Position: between data-link and network layers -Service: best-effort, single hop -Common to both single- vs multiple-hop deployments

9. 9 Properties of SP  SP provides mechanisms for network protocols to operate -Network protocols may introduce policy  Three key elements of SP: -Data Reception -Data Transmission -Neighbor Management

10. 10 Collaborative Interface  Control -ReliabilityBest effort to transmit the msg -UrgencyPriority mechanism  Feedback -CongestionWas the channel busy? Should I slow down? -PhaseWas there a better time to send? Decouple appl sampling from communication

11. 11 Message Reception  Message arrives from link  SP dispatches  Network protocols establish -naming/addressing -filtering SP Receive

12. 12 Message Transmission  Messages placed in shared message pool -All entries are a promise to send a packet in the future  Messages include -Pointer to first packet and # of packets -Control information: reliability and urgency -Feedback information: congestion and phase SP Msg Pool SendReceive

13. 13 Neighbor Management  SP provides a shared neighbor table -Cooperatively managed -SP mediates interaction using table No policy on admission/eviction by SP Scheduling information SP Neighbor Table Msg Pool NeighborsSendReceive

14. 14 SP Data Link AData Link B SP Adaptor ASP Adaptor B Network Protocol 1 PHY APHY B Network Protocol 2 Network Protocol 3 Network Service Manager Link Estimator Link Estimator Neighbor Table Msg Pool NeighborsSendReceive SP Architecture

15. 15 SP Message Pool sp_message_t destination message quantity urgent reliability phase congestion address_t 1 st TOSMsg to send # of pkts to send on or off  adjustment true or false control feedback Neighbor TableMsg Pool Neighbor Table 2 1 NetworkLinkRequiredNeighbor address time on time off listen quality address_t local time node wakes local time node sleeps true or false estimated link quality

16. 16 SP Msg Pool Network Protocol msg* Next Packet Handler packetsSP Message Message Dispatch transmit completed (1) (2) (3)(4) (5) (6) 1 st packet Send inspect Link Protocol SP Message Futures 1)Submit an SP Message for Transmission 2)Message added to message pool 3)SP requests the link transmit the 1 st packet 4)Link tells SP the transmission completed 5)SP asks protocol for next packet 6)Protocol updates packet entry in message pool

17. 17 What SP Isn’t  SP does not dictate any header fields -Messages are opaque to SP  Instead, rely on abstract data types -Can query for address, length, etc  No explicit security mechanism -Message content opaque to SP -Link, Network, and App security can be built transparently to SP

18. 18 Benchmarks  Minimal performance reduction in single hop -Compare to B-MAC paper -Compare to IEEE 802.15.4  Simpler multihop/network protocol code  Power consumption  Network protocol co-existence

19. 19 Results: mica2 Throughput

20. 20 Results: mica2 Throughput

21. 21 Results: mica2 Throughput

22. 22 Results: mica2 Throughput

23. 23 Results: mica2 Throughput

24. 24 Results: Single Hop Benchmarks (802.15.4)

25. 25 Conclusion  SNA: provide context for sharing our community work and accelerate the development and deployment of sensornet applications  Effective link abstraction, SP, allows network protocols to run efficiently on varying power management schemes