Transcript Document

Wireless Networked Sensors
Routing Challenges
Mikhail Nesterenko
In this presentation I used the material from a presentation by
• David Culler, USB http://www.cs.berkeley.edu/~culler/talks/mobihoc.ppt,
http://www.cs.berkeley.edu/~culler/cs294-f03/slides/awoo_oct_2nd_2003.ppt
• Kwong-Don Kang, SUNY, Binghamton www.cs.binghamton.edu/~kang/teaching/cs580s/taming-bvr.ppt
Reading List
•
Deepak Ganesan, Bhaskar Krishnamachari, Alec
Woo, David Culler, Deborah Estrin and Stephen
Wicker, Complex Behavior at Scale: An
Experimental Study of Low-Power Wireless Sensor
Networks , UCLA Computer Science Technical
Report UCLA/CSD-TR 02-0013
• A. Woo and D. Culler. Taming the Underlying
Challenges of Reliable Multihop Routing in Sensor
Networks. In Proc. of the 1st ACM Conf. on
Embedded Networked Sensor Systems (SenSys),
pp 14--27. Los Angeles, Nov 5-7 2003
4/18/2007
WSN
2
Outline
• empirical measurements of low-power
radio performance
 radio neighborhood
 link quality estimation
 issues with simple routing
• mintroute
 link quality estimation
 neighborhood selection
 routing metrics
 simulation results
 experimental results
4/18/2007
WSN
3
Radio Neighborhood
•
radio neighborhood is not
clearly defined
 reception is probabilistic
 not isotropic
• reception rate is low
 “good” link drops 1 out
of 4 packets (cf. ethernet
drops 1 out of 10K)
 changes with time!
4/18/2007
WSN
4
Link Quality
•
three regions based on
reception
 nearly perfect
 unpredictable
 nearly none
• one the fringes some links
are asymmetric
 more than 75% in one
direction
 less than 25% in the
other
• what to do with them?
 detect and ignore?
 embrace?
4/18/2007
WSN
5
Flood-Based Routing Issues
•
simple flood-based
routing is imperfect
• has
 stragglers
 backward links
 dense clusters
4/18/2007
WSN
6
Other Issues
•
large variation in affinity
 asymmetric links
 long, stable high quality links
 short bad ones
• varies with traffic load
 collisions
 distant nodes raise noise floor
 reduce SNR for nearer ones
• many poor “neighbors”
• good ones mostly near, some far
4/18/2007
WSN
7
Outline
• empirical measurements of low-power
radio performance
 radio neighborhood
 link quality estimation
 issues with simple routing
• mintroute
 link quality estimation
 neighborhood selection
 routing metrics
 simulation results
 experimental results
4/18/2007
WSN
8
Link Estimation
•
•
•
4/18/2007
Individual nodes estimate link quality by observing packet success
and loss events
Use the estimated link quality as the cost metric for routing
Good estimator should:
 React quickly to potentially large changes in link quality
 Stable
 Small memory footprint
 Simple, lightweight computation
WSN
9
WMEWMA
•
•
4/18/2007
Snooping
 Track the sequence numbers of the packets from each source to
infer losses
Window mean with EWMA
 WMEWMA(t, a) = (#packets received in t) / max(#packets expected
in t, packets received in t)
 t, a: tuning parameters
 t: #message opportunities
 Take average in a window
 Take EWMA of the average
WSN
10
WMEWA (t =30, a =0.6)
•
4/18/2007
simulation of empirical trace in stable setting
WSN
11
Neighborhood Management
•
•
4/18/2007
Neighborhood table
 Record information about nodes from which it receives packets
 MAC address, routing cost, parent address, child flag, reception (inbound)
link quality, send (outbound) link quality, link estimator data structures
 Propagate back to the neighbors as the outbound rather than inbound link
quality is needed for cost-based routing
 The receiving node may update its own table based on the received
information possibly indicating topology changes Distance-vector based
routing
How does a node determine which nodes it should keep in the table?
 Keep a sufficient number of good neighbors in the table
 Similar to cache management
WSN
12
Management Policies
•
•
4/18/2007
Insertion
 Heard from a non-resident source
 Adaptive down-sampling technique
 Probability of insertion = N/T = neighbor table size / #distinct
neighbors
 At most N messages can be inserted for every T messages
Eviction
 FIFO, Least-Recently Heard, CLOCK, Frequency
WSN
13
#Good neighbors maintainable (table size 40)
•
Frequency Algorithm
 Keep a frequency count for
each entry in the table
 Reinforce a node by
incrementing its count
 A new node will be inserted if
there is an entry with a zero
count
 Otherwise, decrement the count
of all entries and drop the new
candidate
good node – 75% link accuracy
4/18/2007
WSN
14
Background
Link-State and Distance-Vector Routing
•
Link state routing algorithm (ex: DSDV)
 assume knowledge of the network topology and all link costs
 apply Dijkstra’s algorithm to find the shortest path from one source to all
the other nodes
 Implemented via link state broadcast
 memory intensive, has issues with information update
•
Distance vector routing (ex: AODV)
 each node propagates cumulative distance estimator (ex: min # hops) to
all neighbors
 neighbors update their metric and propagate further
 has “counting to infinity” problem
 countered by poisoned reverse or split horizon
4/18/2007
WSN
15
Cost-based routing
•
•
4/18/2007
Key ideas
 Minimize the cost that is abstract measure of distance
 Could be #hops, #retransmissions, etc.
 Minimize # retransmissions: A longer path with fewer #retransmission
could be better than a shorter path with more retransmissions!
Distance-vector based approach implemented by the parent selection
component
 Periodically run parent selection to identify one of the neighbors for
routing
 May also locally broadcast a route message including parent address,
estimated routing cost to the sink, and a list of reception link estimations
of neighbors
 A receiving node may update the neighbor table based on the received
info or drop it
 Flag a child in the table to avoid a cycle
 When a cycle is detected trigger parent selection without the current
parent
WSN
16
Routing Framework
4/18/2007
WSN
17
Underlying Issues
•
•
•
•
4/18/2007
Parent selection
 If connectivity to the current parent is lost, a node disjoins from the tree,
and sets its routing cost to infinity  Reselect a parent
Rate of parent change
 Periodic: Parent selection for every route update msg from neighbors
incurs a domino effect of route changes
Parent snooping
 For example, quickly learn routing info
Cycles
 Monitor forwarding traffic and snoop on the parent address in each
neighbor’s msg -> Identify child nodes and don’t consider them as potential
parents
WSN
18
Underlying Issues
•
•
•
4/18/2007
Duplicate packet elimination
 Use sender address & sequence number
Queue management
 Give priority to originating traffic assuming originating data rate is lower
than forwarding rate
 General fair queuing is not considered in this paper
Relation to link estimation
 Link failure detection based on a fixed number of consecutive xmission
failures can be ineffective over semi-lossy links
 Link quality estimation can be a better judgment of link failure
 Bidirectional link estimations can avoid routing over asymmetric links
 Stability and agility of link estimators can significantly affect routing
 Final tuning must be done while observing its effect on routing
performance
WSN
19
Cost metric
•
MT (Minimum Transmission) metric:
 Expected number of transmissions along the path
 For each link, MT cost is estimated by 1/(Forward link quality) *
1/(Backward link quality)
 Inherently non-linear
 For MT, a substantial noise margin should be used in parent select to
enhance stability
•
Reliability
 Another cost metric
 Product of link qualities along the path
 Not explored in this paper
4/18/2007
WSN
20
Performance Evaluation: Tested Routing Algorithms
•
Minimum Transmission (MT)
 Use the expected #transmissions as the cost metric
 Use a new path if the new cost is lesser by a noise margin
•
MTTM
 Assume a neighbor table can maintain only 20 entries
•
Broadcast
 Root periodically floods the network
 A node chooses a parent that forwards the flooded msg to itself first in
each epoch
 Use the reverse path to reach the root
4/18/2007
WSN
21
Performance Evaluation: Tested Routing Algorithms
•
4/18/2007
Shortest Path
 Conventional distance-vector approach
 Each node picks a minimum hop-count neighbor as the parent and
set its own hop-count to one greater than its parent
 Two variations for performance analysis
 SP: A node is a neighbor if a packet is received from it
 SP(t): A node is a neighbor if its link quality exceeds the
threshold t
o t = 70%: only consider the links in the effective region
o t = 40%: also consider good links in the transitional region
WSN
22
Packet level simulations
•
•
•
4/18/2007
Built a discrete time, event-driven simulator in Matlab
Network of 400 nodes: 20 * 20 grid with 8 feet spacing
Sink is placed at a corner to maximize the network depth
WSN
23
Packet level simulation
Hop Distribution
4/18/2007
Path reliability over distance
WSN
24
Packet level simulation
4/18/2007
WSN
25
Empirical study of a sensor field
•
•
•
•
•
•
4/18/2007
Evaluate SP(40%), SP(70%), MT
50 Berkeley motes inside a building
5 * 10 grid w/ 8 foot spacing
 90% link quality in 8 feet
3 inches above the ground
sink in the middle of short edge of the grid
measurements at night to avoid pedestrian traffic
WSN
26
Link Quality of MT
Hop Distribution
-Vary around 70%
-SP(70) failed to
-SP(70) may suffer
construct a routing
tree
- MT congested: Triple the data origination and route update rate
4/18/2007
WSN
27
Stability
E2E success rate
4/18/2007
WSN
28
Irregular Indoor Network
•
30 nodes scattered around an indoor office of 1000ft2
Link Estimation of a node
to its neighbors over time
E2E Success Rate
4/18/2007
WSN
29
Conclusions
•
•
•
4/18/2007
Link quality estimation and neighborhood management are essential to
reliable routing
 WMEWMA is a simple, memory efficient estimator that reacts
quickly yet relatively stable
MT (Minimum Transmissions) is an effective metric for cost-based
routing
The combinations of these techniques can yield high end-to-end
success rates
WSN
30