The Tenet Architecture for Tiered Sensor Networks
O. Gnawali, B. Greenstein, K-Y. Jang, A. Joki, J. Paek, M. Viera, D. Estrin, R. Govindan, E. Kohler
USC, UCLASenSys 2006
2
The Tenet Two-Tier Architecture
Motes and Masters Multi-node data fusion done on masters Masters program motes using tasks
3
Example Task
Notify application when temperature > 50F
A task contains an arbitrary number of tasklets linked together.
4
Efficiency Costs
Opportunity cost of multi-mote data fusion Motes can still fuse locally-generated data
• Sensor data have high temporal but low spatial redundancy
More data routed to the masters A well-designed WSN will have a small diameter
Higher congestion Application parameters can be tuned, e.g., only high-
confidence pursuers report to masters
5
Five Design Principles Asymmetric Task Communication
Master send mote tasks, mote send master reply, mote cannot initiate tasks (no inter-mote communication)
Addressability Masters can talk to each other, any master can talk to
any mote, a mote can reply to its tasking master Task Library
Each task is a subset of a mote’s generic functionality Robustness
Resilience to extensive network failures Manageability
Tools must offer useful insight into network failures
6
Tenet Task and Task Library Focus on simplicity rather than expressiveness A task is composed of tasklets, which are
parameterized services
Linear composition Tasklets maximize flexibility while remaining
simple Each task has a unique ID, a list of tasklets, and
their parameters Task library composed at compile-time due to
TinyOS
7
Tasklets
Can be composed into a wide range of tasks
8
Task Data Structure
Tasks are dynamically allocated Active Containers hold task data
Cloned when a tasklet repeats
Attributes are 3-tuples:<tag, length, value>
9
The Mote Runtime
Task-aware queues used by services (e.g., wait)
Tenet scheduler operates at tasklet-level granularity Allows multiple tasks to execute concurrently
10
Three Task Operations
Installation Receive a task with a new ID
Modification Receive a task with an existing ID and a body
Deletion Receive a task with an existing ID and no body All active containers associated with a task are
destroyed
11
Example Tasks
Blink
CntToLedsAndRfm
Ping and MeasureHeap
SenseToRfm
12
Data Fusion Example
1. Take 10 samples, timestamp it2. classify as interesting if 3 or more samples >
453. calculates the deviation from the running mean4. displays the sample on the LEDs5. sends the statistic, timestamp, and sample if
interesting
13
Network Subsystem Requirements
Must support different applications on tiered networks
Routing must be robust and scalable Master-to-mote Mote-to-master Small memory footprint
Tasks must be reliably disseminated from any master to all motes
Results must be delivered with end-to-end reliability
14
Addressing and Routing
Every mote and master has a globally unique 16-bit address Motes use TinyOS address Masters use last 16-bits of IP address
Master-to-master: IP routing Mote-to-master: tiered routing
First route to nearest master, then to destination master
Use standard WSN tree-routing protocol like MintRoute
15
Tiered Task Dissemination
Reliably floods tasks to all motes Partial network re-tasking achieved using a predicate
tasklet
Implemented in a generic packet flooding protocol called TRD Reliably floods packets to all nodes (both motes and
masters) Based on beaconing
16
Reliable Transport
Transmits responses from motes to masters Three types
Best effort Reliable transactional Stream transport for high data rate applications
All use hop-by-hop retransmissions The reliable protocols use a simplified version
of TCP
17
Summary of Novel Networking Mech.
18
Evalution: Concurrency
How many tasks can a tmote support at once?
19
Execution Time
Most CPU-intensive tasklet, GatherStatistics, can process 1200 samples in 14.8ms
CPU-bound max sampling rate is 81,000 samples per second
20
Application: PEG
PEG = Pursuit-Evasion Game One or more pursuers collaborate to corral one or
more evaders
Use WSN to help pursuers detect non-line-of-sight evaders
Native implementation uses a leader Multiple nodes sense the evader, leader fuses the
data Stress tests Tenet (no mote-level fusion)
Tenet implementation adjusts the detection threshold
21
PEG Experimental Setup
56 tmotes, 6 stargates Simplifications
Evader detected using RSSI Radio transmit power limited to achieve multihop
• 9-hop diameter One evader, one stationary pursuer on central master
22
PEG Evaluation
Tenet has higher accuracy but higher latency Tenet has lower message overhead
23
Vibration Monitoring Case Study
Tenet used to implement Wisden
•DetectOnSet reduces network traffic•Tenet simplifies programming
24
Manageability
The following task can be used to capture the routing trees:
This can be used to evaluate the task dissemination latency:
25
Robustness
Failure of a master forces routing algorithm to adjust
26
Future Work
Near term Actuation Mote-tier storage Bounded-latency communication
Long term Impact of disconnection due to mobility Authenticity Data Integrity Multi-user control and resource management
27
Conclusion
Tenet simplifies programming while not significantly increasing overhead
28
Application
Pursuit-Evasion Pursuer mobile robots chase after evader robots with
the help of a sensor network Traditional implementation employs mote-tier data
aggregation to reduce redundant evader reports