GARUDA: Achieving Effective Reliability for Downstream Communication in Wireless Sensor Networks Seung-Jong Park, Member, IEEE, Ramanuja Vedantham, Member, IEEE, Raghupathy Sivakumar, Senior Member, IEEE, and Ian F. Akyildiz, Fellow, IEEE Report : Hsiung Chun Kuei IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
GARUDA: Achieving Effective Reliability
for Downstream Communication inWireless Sensor Networks
Seung-Jong Park, Member, IEEE, Ramanuja Vedantham, Member, IEEE,
Raghupathy Sivakumar, Senior Member, IEEE, and Ian F. Akyildiz, Fellow, IEEE
Report : Hsiung Chun Kuei
IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008
http://wshlab2.ee.kuas.edu.tw
Data delivery can be critical
Guaranteed sink-to-sensors
Reliable downstream data
Abstract
2
http://wshlab2.ee.kuas.edu.tw
Outline
3
http://wshlab2.ee.kuas.edu.tw
IntroductionEnergy-aware protocols isn’t enough
Wireless Channel Errors Congestion and Contention Broadcast Storm
4
http://wshlab2.ee.kuas.edu.tw
Delivered reliably Control code Query-data Response result about sensor match data
Assumptions Downstream reliability Communication and node failures 100 % reliable message delivery Message size less then one packet Network model is static
Framework
6
http://wshlab2.ee.kuas.edu.tw
Single-/First-Packet Delivery Benefits
• Robust fading effects and collision.
• Implicit NACK fit in short package.
• Result in low energy.
Framework
7
Framework Wait for first package(WFP) Pulse Transmission
http://wshlab2.ee.kuas.edu.tw8
CS: carrier sensin
http://wshlab2.ee.kuas.edu.tw
Loss Recovery Servers: Core Goal
• Minimize the retransmission overheads.• Constructed in a manage dynamic topology
Rationale of Core• MDS(Minimum Domination Set)• MSC(Minimum Set Cover)
Design Element
9
Framework Instantaneous Core Construction
• Sink– band-ID(bId) = 0
• In 3i bands– Radom wait, and no invite message from the same band. It will be candidate.
– Maintain upstream core’s information
• In 3i+1– S0 is S1’s core ,when the new S0’ core invite again, S1 will trade off each other
by delay time.
• In 3i+2– When time out, the node will sends an anycast “core solicitation message” to
3(i+1) nodes. And then respond after a random waiting delay.
– Boundary condition : not invite form core. Such condition can be detected.
http://wshlab2.ee.kuas.edu.tw10
Framework Loss Recovery for Core Nodes
• Upstream core nodes
• Downstream core nodes– A-map:myBM (successfully received packet),totBM(received and requested
packets)
– If A-map is from a valid source. Updating to totBM.
– Send request , and set expire time. If receive the feedback to update to myBM
– If no response from upstream core, requiring to default upstream core.
• Intermediate noncore nodes
– Set the vFlag to NULL when identifier is equal 3
http://wshlab2.ee.kuas.edu.tw
),,,( vFlagbIdmapACid
),,( bIdmapANCC idid
11
Framework Loss Recovery for Noncore Nodes
• Snoops all (re)transmissions from its core node.
• After Period core presence timer, sends an explicit request to core node that response with A-map
Optimality of the core A-map overhead Number of recovery
events Effect of random wireless
errors
http://wshlab2.ee.kuas.edu.tw15
http://wshlab2.ee.kuas.edu.tw
Performance evaluation
Evaluation of Variants Reliable Delivery within a Subregion
16
http://wshlab2.ee.kuas.edu.tw
Performance evaluation Minimal Set of Sensors
17
http://wshlab2.ee.kuas.edu.tw
Conclusions
Future work With mobility and in the presence of multiple sinks.
We can do .. Take care of core’s energy.
• By reelection
Expand into multimedia• Addition to multi processes.
• How many duplicate does the environment have?
18
Framework –D?Two-Phase Loss Recovery
A-Map(Availability Map)
Function• Loss detection
• Loss recovery
http://wshlab2.ee.kuas.edu.tw20
http://wshlab2.ee.kuas.edu.tw
Performance evaluationSimulation Environment
網路地形• 100 node,650mx650m,randomly deployed• Sink in center• Range 67m• 1Mbps• Message = 100 packets and 25 packets/per second (except for the
single-packet-delivery part)• 1 packet = 1Kbyte
協定參數• MAC protocol : CSMA/CA• Routing : flooding• Simple : 20 randomly topologies• So 95% confidence intervals• Error model : 5% fixed packet loss rate
21
http://wshlab2.ee.kuas.edu.tw
Other reliability semanticsReliable Delivery within a Subregion
• Without loss (100%)• First package decide the core.• Not choose itself?
– 要怎麼決定成為 core? 透過什麼權值來證明它是好機器 ?
Cover the Sensing Field• 2R away from the nearest core node
– Ownership (defined by its transmission range)
• Core node can choose itself as a candidate – 結點少 , 自己判斷成為 core?
Probabilistic Subset Scope sensing(ex:25%) Triggers detected during the preliminary sensing p% be candidate be core
<- 是否使用在對某些興趣點做訂閱時使用 ?
22
http://wshlab2.ee.kuas.edu.tw
Environment considerations• Scarcity of bandwidth and energy.
Message considerations.• The protocol to consider large-sized messages only before.
but WSN need small-sized queries.
• So issues on what kind of loss recovery.
Reliability considerations• 100 percent reliable delivery to only a subregion.
Introduction -D
23
http://wshlab2.ee.kuas.edu.tw
Related work -DBefore
Efficient flooding• Classify: probability-based, area-based and neighbor-knowledge-based• Can’t guarantee the reliability.
“Minimizing Broadcast Latency and Redundancy in Ad Hoc Networks”
• Broadcast tree and schedules transmissions.• Greedy strategy to minimize the latency and the number of retransmissions• Not suit large-scale networks
Pump Slowly, Fetch Quickly (PSFQ) • Relatively slow speed, using in-sequence forwarding.• Recover missing data packets from immediate neighbors.• Single-packet isn’t concider.
TinyDB : Query processor• Minimize power consumption• accuracy of query• No different services
24
http://wshlab2.ee.kuas.edu.tw
ChallengesEnvironment Constraints
Not relying on statically constructed mechanism• dynamics of the network
Tremendous amount of spatial reuse.
Problem definition -D
25
Problem definition -DAcknowledgment (ACK)/NACK Paradox
NACK• Effective loss advertisement mechanism.• Low loss probabilities are not inordinately high.(The package is
small)• Can‘t handle the unique case. When lost message at a part of
node.(The middle node die)• Not aware, it cannot advertise a NACK to request retransmissions.