Download - WSN Performance
-
8/10/2019 WSN Performance
1/20
temas Embebidos 2013/14
SNs - Performance 1
Wireless SensorNetworks PerformanceSistemas Embebidos
2013/14
Pedro Brando
d cc ][
References
These slides are based in chapter 11 ofWireless Sensor Networks: Technology,Protocols, and Applications, KazemSohraby, Daniel Minoli, Taieb Znati, fromWiley-Interscience, ISBN 978-0-471-74300-2,
2007 [WSN-KDT]
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
2
-
8/10/2019 WSN Performance
2/20
temas Embebidos 2013/14
SNs - Performance 2
Background
d cc ][
Basic scenario
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
4
Image from [WSN-KDT]
-
8/10/2019 WSN Performance
3/20
temas Embebidos 2013/14
SNs - Performance 3
d cc ][
Sensor Types
specialized sensing platformE.g.: Spec
generic sensing platformE.g.: SunSPOT
High bandwidth sensingUsing Bluetooth: imote2
gateway-like sensor nodeE.g.: stargate
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
5
d cc ][
Topologies
StructuredE.g.: two-tier Nodes at low level onlysense, upper level senseand relayMultihop
Routing more efficientMore energy for topologyformationMore node complexity tohandle the protocol
Ad Hocflat and unstructuredtopologyAll nodes sense andrelay
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
6
-
8/10/2019 WSN Performance
4/20
temas Embebidos 2013/14
SNs - Performance 4
d cc ][
Topologies
Multihop multipath, route redundancyBattery depletion
dysfunctional node
Disjoint topology (when dysfunctionalnodes reach threshold)
Mobility causes similar problems
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
7
d cc ][
Traffic
From sensor nodes to a single sink Many to one (convergecast)
May be for several sinks multipaths
Nodes closer to sink will have more trafficto handle
Dysfunctional sooner
deploy more nodes around the sink, rotatenodes
Data aggregation
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
8
-
8/10/2019 WSN Performance
5/20
temas Embebidos 2013/14
SNs - Performance 5
d cc ][
WSN Service type
basic service: detect certain events andreport them.
data is usually small (a few bytes or a fewbits)
transmit more than one event in a single dataunit if application reporting frequency allows it.
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
9
d cc ][
Design Factors for WSNs
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
10
Table from [WSN-KDT]
-
8/10/2019 WSN Performance
6/20
temas Embebidos 2013/14
SNs - Performance 6
MAC protocols
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
11
d cc ][
MAC protocols energy waste
packet collision require retransmissionenergy waste
Wireless shared medium
Control messages for data transmission (e.g.,RTS/CTS) consume energyOverhearing and idle listening energywaste.
Overhearing: node receives packets destined forothers.Idle listening: nodes listen on the channel to get itsstatus
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
12
-
8/10/2019 WSN Performance
7/20
temas Embebidos 2013/14
SNs - Performance 7
d cc ][
Some approaches
S-MAC: low-duty-cycle coordinated sleepand wakeup time periods to reducepower consumption while achieving highthroughput.
LEACH: recall that it uses TDMA in clusterand CDMA between clusters
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
13
d cc ][
Measuring MAC performance
Collision probability
Control overheadDelay
Throughput.
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
14
-
8/10/2019 WSN Performance
8/20
temas Embebidos 2013/14
SNs - Performance 8
Routing protocols
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
15
d cc ][
Classification
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
16
Image from [WSN-KDT]
-
8/10/2019 WSN Performance
9/20
temas Embebidos 2013/14
SNs - Performance 9
d cc ][
Data centric routing advantage
Easy deploymentData aggregation enables energy saving
Low overhead in control messages
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
17
d cc ][
Measuring routing performance
Computational overhead
Communications overheadPath reliability
Path length
Convergence rateStability
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
18
-
8/10/2019 WSN Performance
10/20
temas Embebidos 2013/14
SNs - Performance 10
Transport protocols
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
19
d cc ][
Main factors
Congestion controlReliability
Different from normal netsReceive info from at least one node inselected area
Receive certain ratio of successful transmissionsfrom a node
Have a hop-by-hop mechanism forcongestion control and loss recoveryFairness
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
20
-
8/10/2019 WSN Performance
11/20
temas Embebidos 2013/14
SNs - Performance 11
Performance models
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
21
d cc ][
MetricsSystem lifetime:
duration of time until some node depletes all itsenergy;
duration of time until the QoS of applicationscannot be guaranteed
duration of time until the network has been
disjoined.Energy efficiency: number of packets thatcan be transmitted successfully using aunit of energy.
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
22
-
8/10/2019 WSN Performance
12/20
temas Embebidos 2013/14
SNs - Performance 12
d cc ][
Metrics (cont.)
Reliability: can be defined as the ratio ofsuccessfully received packets over thetotal number of packets transmitted.Coverage: ratio of the monitored space tothe entire space.Connectivity: evaluate how well thenetwork is connected and/or how manynodes have been isolated.QoS metrics: applications may have QoSrequirements such as delay, loss ratio, andbandwidth.
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
23
d cc ][
Traffic model
Event-Based Delivery:Monitor continuously
Detect event and transmit (with possible value)
Continuous DeliveryContinuously or periodically
Query-Based DeliverySink issues query
Hybrid DeliveryDiverse types of data
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
24
-
8/10/2019 WSN Performance
13/20
temas Embebidos 2013/14
SNs - Performance 13
d cc ][
Energy
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
25
Image from [WSN-KDT],from a paper from 2002
d cc ][
Energy models
Model for Sensing:Energy expenditure proportional to (radius) 2 or(radius) 4 of area
Model for Computation:Low energy per clock cycle energy efficiencyEnergy efficiency: energy per instruction
ATMega 128L at 4 MHzConsumption: 16.5mW,Efficiency: 242 MIPS/W or 4 nJ/instruction
ARM Thumb at 4 MHz:Consumption: 75mWEfficiency: 480 MIPS/W or 2.1 nJ/instruction
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
26
-
8/10/2019 WSN Performance
14/20
temas Embebidos 2013/14
SNs - Performance 14
d cc ][
Energy models
Model for CommunicationOptimal distance:
= 2
( )2 , where
Ec: base energy required to run thetransmitter/receiver circuitry
e: unit energy required for the transmitter amplifier, itis different for distances bigger than the crossoverdistance
s: power factor 2 or 4 if distance is smaller or greaterthan crossover distance respectively
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
27
d cc ][
Energy models
Node ModelActive (A) and Sleep (S) states
If A:R state: receive and send data
N state: send only if in backlog
Next hop node:Wait (W): either in S or N
Forwarding (F): in R
Use Markov chain model
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
28
-
8/10/2019 WSN Performance
15/20
temas Embebidos 2013/14
SNs - Performance 15
d cc ][
MAC Model
Gupta and Kumar define successfultransmission from a to b with max radiorange of r if:
,
, > where k is any other node receiving
, > where l is any other node transmitting
From this an interference model was
derived based on the CSMA/CA of 802.11DCF
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
29
d cc ][
Routing model
Energy consumed for a generic path P
= , ( , )
next hop of i in path P
, ( , ) : energy from node i to node assuming data size is bits and distance
Results:
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
30
-
8/10/2019 WSN Performance
16/20
temas Embebidos 2013/14
SNs - Performance 16
d cc ][
System model
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
31
Closed loop model
Case analysis
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
32
-
8/10/2019 WSN Performance
17/20
-
8/10/2019 WSN Performance
18/20
temas Embebidos 2013/14
SNs - Performance 18
d cc ][
Analysis
Total nodes
: bit error rate and packet error rate: = 1 1
Average retransmission for K maxretransmissions
For K = =
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
35
d cc ][
Analysis (cont.)
Total energy per hop = + +
Converged data rate
Where a is aggregation efficiency
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
36
-
8/10/2019 WSN Performance
19/20
temas Embebidos 2013/14
SNs - Performance 19
d cc ][
Analysis (cont.)
System lifetime
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
37
d cc ][
Max-min fairness
See book
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
38
-
8/10/2019 WSN Performance
20/20
temas Embebidos 2013/14
The endfor WSN Performance
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
39
d cc ][
Summary
Performance and traffic managementSensors
MAC
Routing
Transport
Energy Models
Analysis
SE 2013/14 - WSNs 2 - pbrandao (see at beginning)
40