Title: PHD DISSERTATION DEFENSE ReceiverCost Cognizant Maximal Lifetime Routing in Embedded Networks: Model
1PHD DISSERTATION DEFENSEReceiver-Cost Cognizant
Maximal Lifetime Routing in Embedded Networks
Model and Solutions
- Guofeng Deng
- Advised by Dr. Sandeep Gupta
- The IMPACT Lab
- Ira A. Fulton School of Engineering
- Arizona State University, Tempe
2Outline
- Motivation
- Intrinsic energy constrains in wireless ad hoc
networks (WANET) - Existing research assumed that power for transmit
dominates, and ignored power for receiving - For low power wireless devices, which are used
widely nowadays, however, receiving power is not
negligible. - Background
- WANET overview and the intelligent shipping
container project - Receiving energy cost models
- Key Results
- Receiver-cost cognizant maximal lifetime routing
- Maximal lifetime routing in mobile networks
- Conclusion Future Work
3Highlights
- What it is all about
- investigating the impact of receiving energy
costs by revisiting the maximal multicast
lifetime routing problem, which is trivial if
receiving packets is free - Key contributions
- Receiving costs change the problem dramatically
- NP-hard if receiving power is adaptive to
received signal strength Deng, Gupta,
Varsamopoulos, IEEE Comm. Letters 2008 - NP-hard if nodes consume energy for overhearing
as well Deng Gupta, ICDCN'06 - Handling receiving costs properly improve
multicast lifetime compared with disregarding
them - By 15 with no overhearing costs Deng Gupta,
Globecom06 - By 60 with overhearing costs Deng Gupta,
ICDCN'06 - First distributed algorithm to adapt to node
mobility for maximal multicast lifetime Deng,
Mukherjee Gupta, in preparation
4Wireless Ad Hoc Network Overview
- WANET
- Distributed networks nodes talk to others in the
proximity directly - Wireless devices low power, small form factor
devices with short comm. range - Multicast medium one transmission can reach
multi-entities in proximity - Features
- Flexible robust no dependency on fixed
infrastructure - Scalable can accommodate large number of
entities - Low cost, low maintenance, mobile,
- Applications
- Surveillance and rescue environment monitoring,
Body Area Network (BAN), - Challenges
- Limited resources (e.g. energy, bandwidth)
5WANET Application Intelligent Shipping
Container (ISC)
- Background Global nature of todays economy
- 90 of the worlds trade is transported in cargo
containers - 10 million cargo containers enter U.S. ports each
year - Motivations
- Homeland security only 5 can be inspected
because of todays limited time and money - Commercial values lack of end-to-end visibility
for supply chain and chain of custody - Goal
- Architecture design that meets various
requirements - Verification of currently available technologies
- A joint effort between Intel Inc.
- the Impact Lab.
6ISC Hierarchical Container Network
- Internal container network (within each
container) - Wireless Sensor Network (WSN) collecting
environmental parameters - Radio Frequency Identification (RFID) automatic
and unique identification, multi-level tracking
(e.g. products, packages, pallets, containers,
etc) - Gateway point of access from outside container,
on scene data processing and storage - External container network
- WANET interact with neighboring containers
7ISC Severe Energy Constraints
- MicaZ mote with MTS310 Sensor board
- Broadcasts a packet every 10 sec with its voltage
level - Uses the power saving mode (switching off radio
and sensor board after readings) - 2 new AA batteries
- The base station (4 meters away) collects packets
- The mote lasts about 46 days
- Had to use a car battery to power Stargate
(gateway) and RFID reader for a 5-day shipment.
- Government regulation requires lifetime at least
1 year.
8Motivation Receiving Power is Not Negligible
- Most existing research aimed to conserve energy
for transmitting packets and neglected energy
consumed for receiving packets - Receiving power is not negligible in low power
devices - Chipcon CC2420 (single-chip 2.4 GHz IEEE 802.15.4
compliant RF transceiver) data sheet
9Motivation Receiving Power Affects Route
Optimality
- Example scenario to transmit same packets from 1
to 2 and 3, i.e. multicast - Goal to transmit as many packets as possible
before any node exhausts its battery - Assumption identical nodes, B is battery
capacity, a link is associated with energy
consumed for transmitting a packet over the link.
1
1
4
3
3
2
2
3
3
2
Energy cost for rcv a pkt Total
number of packets rcved by 2 or 3
0 B/3
(node 1 dies first) B/4 (node 1 dies first)
4 B/6
(node 2 dies first) B/4 (all nodes die at same
time)
10Receiving Energy Cost Characterization
- Media Access Control protocol
TDMA based a node may switch off the transceiver
based on some schedule, avoid overhearing
irrelevant packets. But it will consume energy
for receiving packets designated to it.
Random access a node may overhear transmissions
in the proximity and consume energy for
demodulating signals not interested in.
2
2
1
1
3
3
11Receiving Energy Cost Characterization
Regular decoder energy consumed for decoding a
signal is independent on signal strength, i.e.
constant
Turbo decoder energy consumed for decoding a
signal is inversely proportional to signal
strength, i.e. adaptive
Based on transmitter-receiver power tradeoff
Vasudevan et al. Infocom'06
12Receiving Energy Cost Models
Objective to investigate maximal multicast
lifetime problems under each of these receiving
energy cost models.
13Maximal Multicast Lifetime Routing in WANET
- Multicast traffic
- A single source generates multicast packets
- A set of nodes in the network need to receive the
packets - Metrics
- The duration until the first node in the network
to fail due to exhausted battery - State of the art
- Solvable in polynomial time when the receiving
energy cost is 0
14Related work
- Energy efficient multicast routing
- Reduce overall energy consumption for multicast
traffic - Take advantage of multicast media
- Maximal lifetime multicast routing
- Extend the duration until the occurrence of some
application dependent critical events - Balance energy consumption among nodes
- Static vs. dynamic approaches
- Overhearing energy costs
- Studied for data-gathering routing
- Adaptive receiving costs
- Studied for data-gathering routing
15MaxMLT under DCR
- Problem
- Maximizing multicast lifetime under the DCR
model, i.e. designated constant receiving energy
costs - Designated Receiving Power algorithm (DRP)
- In the directed network graph, there is a link
(u,v) if u can be received by v when peak
transmit power is used. - Convert the network graph to so called INverse
longevity Graph (ING) - Run Prims algorithm on the ING to generate a
multicast tree - Result
- Optimal solution of time complexity O(n2 log n),
where n is the number of nodes
16MaxMLT-DCR Simulation Results
RX peak TX for each node u
For each node u, select RX randomly between peak
TX and 2X peak TX
ZRP Zero Receiving Power
Network size (density) number of nodes in the
network
All the nodes are destinations and have identical
battery capacity and peak TX power.
17MaxMLT under OCR
- Problem
- Maximizing multicast lifetime under OCR, i.e.
overhearing constant receiving energy cost - Challenges
- NP-hardness reduce set cover to MaxMLT under OCR
18MaxMLT under OCR NP-hard
Source node
Forwarding nodes
Destination nodes
Set cover
MaxMLT under OCR
- Observations
- Node s will die first
- Lifetime of resulting multicast tree is
determined by the number of forwarders
- Assumptions
- identical battery capacity
- all links are associated with same transmit power
19MaxMLT-OCR Heuristic Solution
PRP Proximity Receiving Power algorithm
- Link weight computation with various metrics
adding link (2,4) to the existing tree.
on-tree node
1
Note Link metric defines how the receiving power
is taken into account.
non-on-tree node
4
link being considered
link that has been established
transmit costs taken into account
receiving costs taken into account
CRP Cumulative Receiving Power algorithm
(extending DRP for comparison)
overhearing costs taken into account
PRP
ZRP
DRP
CRP
4
1
4
4
4
1
1
1
2
2
2
2
3
3
3
3
20MaxMLT-OCR Simulation Results?
RX peak TX
RX 2X peak TX
Identical battery capacity and peak TX power
21MaxMLT-OCR A Mcast Tree Snapshot
- The source node is surrounded by a hexagram and
the rest are destinations - Solid lines constitute mcast trees
- A circle represents the transmit range of the
node in the centre - The diameter of a solid grey dot represents the
magnitude of overhearing costs
- Observation PRP tends to increase transmit power
and reduce num of transmitters to decrease
overhearing costs.
22MaxMLT under DAR
- Problem
- Maximizing the multicast lifetime under the DAR
model, i.e. designated adaptive receiving costs. - Assuming discrete levels of transmitting and
receiving power - State of the art
- A binary search optimal solution to a weaker
problem, in which the multicast tree structure is
given - Challenges
- NP-hardness
23MaxMLT-DAR Chain of Transforms
- A chain of reduction from X3C (exact cover by
3-sets) to MaxMLT under DAR
Is there any m-arbor in k-subgraph?
Decision problem of MAL to seek a m-arbor whose
lifetime is no less than some positive bound
Maximal m-arbor lifetime m-arbor is a tree
defined in an auxiliary graph any m-arbor can be
mapped to a mcast tree in the original graph with
same lifetime and vice versa
Special case of MaxMLT nodes can adjust transmit
and receive power only in discrete levels. We
also assume identical battery capacity.
24MaxMLT-DAR Chain of Transforms contd
A WANET and its auxiliary graph. Each node has
two transmit levels and two receive levels.
25MaxMLT-DAR Chain of Transforms contd
Reducing X3C to AIK in a k-subgraph. The transmit
(receive) vertices in each bipartite are sorted
in ascending order using represented
transmit/receive power levels.
A flat through path go through each bipartite
once and only one
26MaxMLT in Mobile Ad Hoc Networks
- Problem
- Maximizing multicast lifetime when nodes are
mobile - Challenges
- Dynamic networks node mobility and residual
energy changes - No distributed solutions exist
- Solution
- MSL (Multicast Service Lifetime)
- Multicast lifetime definition suitable for mobile
networks - DAMIL (Distributed and Adaptive Multicast
Lifetime algorithm) - Propose a new metric to decentralize routing
decisions - Distributed and adaptive solution that adopts the
new metric
27MaxMLT-MANET MSL
- Quality of multicast service
- num of packets received (vs. time)
- by some or all destinations (address fairness
among destinations) - in some period of time (adapt to dynamic networks
in timely manner) - MSL is a measurement of quality multicast service
received
28MaxMLT-MANET ?Max-Lifetime Tree
- A new metric that leads to distributed algorithm
- Link weight
- Node weight
- s,i is a path from s to i (s-path) and Wmax is
a large value - Optimality a multicast tree in which each node
maximizes its weight is a maximum-lifetime
multicast tree
29MaxMLT-MANET DAMIL
- Data structure a status table contains an entry
for each neighboring node and itself - wi node weight
- pi parent id
- hi hop-count (distance from the source)
- fi forwarding control boolean (FCB, whether to
forward packets to some children) - Periodic beacons (Control info is carried in
periodic beacons) - (wi, pi, hi, fi )
- Activities
- Each node repeatedly seeks the s-path that
maximizes its weight - Upon receiving a beacon
- An entry is created if the sender is a new
neighbour - Build or refine s-paths for gain in node-weight
- Updates the FCB accordingly
30MaxMLT-MANET Example
Node c moved to a new location. Assume symmetric
link weight and Wmax 99.
31MaxMLT-MANET Simulation
- Measurement
- The quality of a multicast service is denoted by
Q(?,G,?), where ?,G,and ? represent window size,
destination threshold and data threshold
respectively. - the MSL is defined as the period of time until
the service quality drops below Q - Comparison algorithms
- WMST updates a maximum lifetime tree
periodically outperforms most lifetime
maximizing protocols in static networks - SS-SPST-E a distributed energy minimizing
multicast protocol designed to overcome the
impact of node mobility.
32MaxMLT-MANET Simulation Results
50 nodes totally, D is a set of destinations,
source generates four 512B packets per second.
33MaxMLT-MANET Simulation Results
34Conclusion Future Works
- Conclusion
- Showed receiving costs change the maximal
multicast lifetime problem dramatically - Showed handling receiving costs properly improves
multicast lifetime significantly compared with
disregarding them - Proposed first distributed algorithm to maximize
multicast lifetime in mobile ad hoc networks - Future works
- Scavenging energy management
35Publication
- G. Deng, S. K. S. Gupta, and G. Varsamopoulos,
Maximizing Multicast Lifetime with
Transmitter-Receiver Power Tradeoff is NP-Hard,
IEEE Communications Letters, Vol. 12, No. 9,
September 2008 - G. Deng and S. K. S. Gupta, On Maximizing Network
Lifetime of Broadcast in WANETs under an
Overhearing Cost Model, ICDCN 2006, LNCS 4308 - G. Deng and S. K. S. Gupta, Maximizing Broadcast
Tree Lifetime in Wireless Ad Hoc Networks, IEEE
GLOBECOM'06, San Francisco, CA - Deng, G. and Gupta, S. K. S. (2005). Maximizing
multicast lifetime in wireless ad hoc networks.
In L. T. Yang M. Guo (Eds.), High-Performance
Computing Paradigm and infrastructure (pp.
643-660). Hoboken, NJ John Wiley Sons. - S. J. Kim, G. Deng, S. K. S. Gupta and M.
Murphy-Hoye, Enhancing Cargo Container Security
during Transportation A Mesh Networking Based
Approach, IEEE HST, Waltham, MA, USA, April 2008.
- S. J. Kim, G. Deng, S. K. S. Gupta and Mary
Murphy-Hoye, Intelligent Networked Containers for
Enhancing Global Supply Chain Security and
Enabling New Commercial Value, COMSWARE,
Bangalore, India, 2008. - G. Deng, T. Mukherjee, and S. K. S. Gupta, DAMIL
A Distributed and Adaptive Algorithms to Extend
Multicast Service Lifetime in MANETs, in
preparation for IEEE Communications Letters.
36Thank You!
37Motivation Survive Energy Constrains
- Replenish battery
- Battery replacing expensive and impractical for
large scale networks, such as sensor networks - Explore ambient energy source limited capability
- Consume energy intelligently
- Energy efficiency reduce total amount of energy
consumed - Lifetime enlarge network life span as a whole
38Intelligent Shipping Container contd
INTER-Container TelosB mote Attached to nearby
containers. Proximity motes form an ad hoc
(multi-hop) inter-container network.
GPS Receiver 1
Containers
MICAz mote
External Hosts
Stargate
Internal Wireless Sensor Networks
USB Memory Card
MICAz mote 2.4 GHz
2.4 GHz
51-pin
USB
Stargate Managing Internal network (hardware,
power and security) data processing, routing
outgoing packets to external interface.
Ethernet
Mobile Computing Computers at point of work
(Handhelds) at the Data Center. Held by custom
officers and load/unload workers. Querying
current and historical data and DB downloading
from the logging systems.
Enterprise Servers Computers at the Data
Center. Collecting real-time data from
containers, managing DB responding to critical
events reported by containers.
802.11
RS232
PCMCIA
Compact Flash
GPRS PCMCIA Modem
802.11 Compact Flash card
Cellular Network
Arch Rock Edge Server Linux computer running web
services-based environment with web UI for setup,
control, monitor, management of diverse
wireless sensor networks.
Arch Rock DataLogger Low-power Embedded Linux
computer running local data collection
management of diverse wireless sensor networks
Ethernet
39ISC Hierarchical Container Network
- External Container Network
- A container forms and participates in networks
with their neighbors dynamically.
- Internal Container Network
- The network inside a container is isolated from
the dynamic changes outside a container.
40ISC Network Entities
PDA monitor and manually control Stargate, e.g.
start/stop RFID reader
Skyetek M8 RFID reader Cushcraft antenna
MICAz mote Read RFID tags and forward the
reading via wireless interface
Stargate MICAz mote WiFi card memory
card data collection and processing, database
management.
MICAz motes MTS310 environmental sensor
TelosB motes with onboard sensors environmental
sensor
Base station startup control and monitoring
41ISC System Tests
- Tested in a standalone container over several
months in Chandler, Arizona, US - Tested in a container yard in a 33 stacked
container configuration in South Kearny, New
Jersey, US - Tested during a 5-day shipment from Singapore to
Kaohsiung, Taiwan
42Background Generic Receiving Power Form
- receiving energy of node v
- transmit energy of node u
- base receiving energy per bit of v
- monotonic non-increasing adaptive receiving
energy function ranging from 0 to 1 - transmission rate of i (bps)
- min transmit power of node i to reach node v
- distance between nodes u and v
- fading exponent
- integer parameters that can be either 0 or 1
For example, under OCR, if for any
i and j, then
43Background Designated Receiving Cost
transmitter
1
designated receiver
2
2
not related to the transmitter
1
3
transmit power
3
receiving power
44MaxMLT-OCR PRP A Heuristic Solution
- Take into account the effects of overhearing
explicitly on both transmitter and receiver - The weight of link (i,j) inverse link longevity
-- incorporates the overhearing cost of i caused
by v it also takes into account the overhearing
costs of v and u due to adding link (i,j) . - Run Prims algorithm to generate a tree that
minimizes the maximum link weight
v
Transmission link Overhearing link
u
k
i
j
45Background Overhearing Cost
transmitter
1
designated receiver
2
2
not related to the transmitter
1
3
transmit power
3
receiving power
46Background Constant Receiving Cost
transmitter
1
receiver
2
2
1
transmit power
receiving power
47Background Adaptive Receiving Cost
transmitter
1
receiver
2
2
1
transmit power
receiving power
48Background Adaptive Receiving Cost contd
Based on transmitter-receiver power tradeoff
Vasudevan et al. Infocom'06
transmitter
1
receiver
2
2
1
transmit power
receiving power
49MaxMLT-OCR? Feasible Metrics
- Proposed metric Proximity Receiving Power (PRP)
- Take into account transmit power cumulated
receiving power of the transmitter, receiving
power of the receiver, transmit power cumulated
receiving power of all the affected neighbors - Possible metrics
- Zero Receiving Power (ZRP)
- transmission power only, i.e., assume 0
reception cost - Designated Receiving Power (DRP)
- transmitter's transmit power, receiver's
receiving power - Cumulative Reception Power (CRP)
- transmit power cumulated receiving power of
the transmitter, receiving power of the receiver
50MaxMLT-DCR Solution Analysis
- Optimal analysis
- Result
- Optimal solution of time complexity O(n2 log n),
where n is the number of nodes
51ISC Hierarchical Network Structure
- Server
- At shippers control center
- Communication with gateways via the External
Container Network - External Container Network
- To support the communication between gateways and
interface between the server and a gateway - Internal Container Network
- To support the communication between devices
within a container (e.g. a gateway, a RFID
reader, and sensors)