1. Collection Tree Protocol
ACM SenSys’09
Slides revised from Omprakash Gnawali

2. Outline Introduce Control Plane (two principles of CTP) Data Plane
Datapath validation
Adaptive beacons
Data Plane
Evaluation

3. Data collection protocols
Three phases in handling sensory data in IoT:
Sensory data acquisition
Sensory data collection
Sensory data computation
CTP is a kind of data collection protocol
Base Requirement
Reliability: > 90% of packet
Robustness: Operate without tuning or configuration
Efficiency: Use minimum of transmissions
Hardware Independence: no specific chip feature

4. CTP
sink
Is a protocol that computes routes from sensor to one or more sinks(base stations)
Build and maintains minimum cost tree(s) with the sink(s) as root
A distance vector protocol

5. Parent Selection Minimize transmission costs(ETX)b
Minimize transmission costs(ETX)
Every node maintains an estimate of the costs of a route to sink 10 e 3 1 2 d a
sink 4 3 2 2 2 3 f c
e: EXT(e) = 1
f: EXT(f) = 2
d: 2 + EXT(e) < 2 + ETX(f)
b: 10 + EXT(e)
c: 3 + EXT(f)
a: 4 + EXT(d) < 3+ EXT(b) and 2 + EXT(c)

6. Challenges for data collection
Link dynamics
Links can be highly dynamic, particularly in the 2.4 GHz frequency space
Wireless links can change in the order of a few hundred milliseconds
Routing inconsistencies
Rapid topology change can cause routing inconsistencies and route loops, cause packet loss in the data plane

7. Common Architecture Control Plane Data Plane Routing Engine
Application
Fwd
Table
Link Estimator
Forwarding Engine
Link Estimator: estimate and compute ETX to neighbors
Routing Engine: compute route and cost to sink, generate fwd table
Forwarding Engine: forward data packets according to fwd table
Link Layer

8. CTP Noe Enable control and data plane interaction Control Plane
Two mechanisms for efficient and agile topology maintenance
Datapath validation
Adaptive beaconing
Routing Engine
Application
Link Estimator
Forwarding Engine
Link Layer

9. Outline Introduce Control Plane(two principles of CTP) Data Plane
Datapath validation
Adaptive beacons
Data Plane
Evaluation

10. Routing Consistency Next hop should be closer to the destination
Maintain this consistency criteria on a path
Inconsistency due to stale state ni
ni+1 nk

11. Routing InconsistencyX
Cost does not decrease
Route inconsistency can cause routing loops
8.1
3.2 C B
4.6 A D
6.3
5.8

12. Datapath validation Use data packets to validate the topology
Inconsistencies
Loops
Receiver checks for consistency on each hop
Transmitter’s cost is in the data packet header

13. Datapath validation X Datapath validation Inconsistency 8.1 3.2
Cost in the packet
Receiver checks
Inconsistency
Larger cost than on the packet
On Inconsistency
Don’t drop the packets
One beacon interval pause in forwarding
Immediately Signal the control plane X
3.2 < 4.6?
8.1 < 4.6?
8.1
8.1
3.2 C
4.6<5.8?
4.6 < 6.3?
4.6
4.6 B
4.6
6.3
5.8 < 8.1? A
5.8 D
6.3
5.8

14. Outline Introduce Control Plane(two principles of CTP) Data Plane
Datapath validation
Adaptive beacons
Data Plane
Evaluation

16. How Fast to Send Control Beacons?
Control beacons are broadcasts
Other protocols typically use periodic beacons to update network topology and link estimates
Using a fixed rate beacon interval
Faster rate lead to higher cost
Slower rates lead to less agility
Agility-efficiency tradeoff
Agile+Efficient possible?
CTP use Trickle Algorithm[Levis, 2004]
Fast rate: more bandwidth and more energy, inference data packets
Slow rate: less agility. if topology changes, cannot update immediately, route inconsistence and route loop can exist for a long time

17. Control Traffic Timing
In CTP:
Start with lowest intervals of 64ms
When interval expires double it up to 1 hour
Reset to lowest interval when inconsistent
Reset the interval on three events
Receive packet ETX <= self ETX
Its routing cost decreases Significantly
It receives a packet with P bit set TX
Increasing interval Reset interval

18. Adaptive vs Periodic Beacons
Total beacons / node
1.87 beacon/s
0.65 beacon/s
Time (mins)
Tutornet
Less overhead compared to 30s-periodic

19. Path established in < 1s Efficient and agile at the same time
Node Discovery
A new node introduced
Total Beacons
Path established in < 1s
Tutornet
Time (mins)
Efficient and agile at the same time

20. Outline Introduce Control Plane(two principles of CTP) Data Plane
Datapath validation
Adaptive beacons
Data Plane
Evaluation

21. Data Plane Design Queueing discipline Per-client queueing [top-level]
each client may have one outstanding packet
achieves better fairness than a shared queue
Hybrid send queue [lower-level]
contains both route-through and locally-generated traffic
duplicate packets are dropped [i.e., not inserted in the queue]

22. Data Plane Design Transmit Timer Self-interference between packets
Rate Control: delay the transmission of packets: randomized between (1.5, 2.5) packet times

23. Data Plane Design Transmission cache
For each transmitted packet insert(origin, seqNo, THL)
Determine if a packet is duplicate

24. Outline Introduce Control Plane(two principles of CTP) Data Plane
Datapath validation
Adaptive beacons
Data Plane
Evaluation

25. Variations in hardware, software, RF environment, and topology
Experiments
12 testbeds
nodes
7 hardware platforms
4 radio technologies
6 link layers
Variations in hardware, software, RF environment, and topology

26. Reliable, Efficient, and Robust
Testbed
Delivery Ratio
Wymanpark
0.9999
Vinelab
Tutornet
NetEye
Kansei
0.9998
Mirage-MicaZ
Quanto
0.9995
Blaze
0.9990
Twist-Tmote
0.9929
Mirage-Mica2dot
0.9895
Twist-eyesIFXv2
0.9836
Motelab
0.9607
Motelab has slightly low delivery ratio, packet loss is the dominant cause of failure. Because Motelab nodes are only intermittently and sparsely connected
False ack
Retransmit
High end-to-end delivery ratio (but not on all the testbeds!)

27. Reliable, Efficient, and Robust
Reliable: Packets delivered to the sink

28. Reliable, Efficient, and Robust
Efficient? TX required per packet delivery
Tutornet
CTP Noe
Low data and control cost

29. Reliable, Efficient, and Robust
Time (mins)
Delivery Ratio
10 out of 56 nodes removed at t=60 mins
Robust? Performance with disruption
Tutornet
No disruption in packet delivery

30. Reliable, Efficient, and Robust
Nodes reboot every 5 mins
Routing Beacons
~ 5 min
Tutornet
Delivery Ratio > 0.99
High delivery ratio despite serious network-wide disruption (most loss due to reboot while buffering packet)

31. Summary of Results 90-99.9% delivery ratio Compared to MultihopLQI
Testbeds, configurations, link layers
Compared to MultihopLQI
29% lower data delivery cost
73% fewer routing beacons
99.8% lower loop detection latency
Robust against disruption
Cause for packet loss vary across testbeds

32. Thank you!