garuda: achieving effective reliability for downstream communication in wireless sensor networks...
TRANSCRIPT
GARUDA: Achieving Effective Reliability forDownstream Communication in Wireless Sensor Networks
Seung-Jong Park et al
IEEE Transactions on mobile computingFeb, 2008
presented by jae-hong Kim
2 / 31
Contents
Transport Layer Issues for ad hoc WSN Reliable bi-directional transport protocol
Characteristics of GARUDA Pulsing based solution Virtual infrastructure called core Two-Phase Loss Recovery Multiple Reliability Semantics
Evaluation Discussion
3 / 31
Transport Layer Issues for ad hoc WSN
Vision Statement Reliable and Robust bi-directional (sink to
sensors and sensors to sink) transport pro-tocol for Ad-hoc Wireless Sensor Networks
4 / 31
To the knowledge …
Up to this point Reliability and Robustness has been ignored;
Possible reason: WSN is low-cost; Not necessary (due to redundant data) And also difficult
But … We require reliability …
Disaster Recovery Military Applications etc
5 / 31
Focus
To achieve reliability Reliable Transport Layer No packet loss Bi-directional Reliability
Figure from Akyildiz et al, “Wireless Sensor Networks: A Survey”, Computer Networks, 38(4):393-422, 2002.
6 / 31
Is it challenging?
Limitations of sensor nodes Application specific requirements
Objectives Reliable Transport Flow Control Congestion Control Self Configuration Energy Awareness
7 / 31
Types of Data
Single Packet Block of packets Stream of Packets
8 / 31
Today’s Situation
Downstream Reliability: from sink to Sensors
Reliability semantics are different PSFQ (Block of packets data) MOAP (block of packets data) GARUDA (Block of packets data) (Single Packet)
9 / 31
Introduction
Reliable downstream point-to-multipoint data deliv-ery The need for the reliability is dependent on the type of
applications. Ex) security application
Reliability in multihop wireless networks vs Reliabil-ity in wireless sensor networks Environment considerations
Limited life time, bandwidth, energy, size of the network Message considerations
In a sensor networks, small-sized queries Reliability considerations
Dependent on reliability semantics
10 / 31
GARUDA
GARUDA is a large mythical bird or bird-like creature that ap-pears in both Hindu and Buddhist mythology
Transport reliably
11 / 31
Characteristics of GARUDA
1. An efficient pulsing-based solution for reli-able short message delivery
2. A virtual infrastructure called the core, which approximates an optimal assign-ment of local designated servers
3. A two-stage negative acknowledgment (NACK) based recovery process and out-of-sequence forwarding
4. A simple candidacy based solution to sup-port the different notions of reliability
12 / 31
ACK / NACK Paradox (1)
NACKs Well established as an effective loss adver-
tisement in multi-hop wireless networks In case loss probabilities are not inordinately
high Not for single-packet delivery or all packets
are lost It cannot possibly advertise a NACK to request
retransmissions
13 / 31
ACK / NACK Paradox (2)
ACK implosion
14 / 31
Pulsing based solution (1)
It incorporates an efficient pulsing based solution, which informs the sensor nodes about an impending reliable short-message delivery by transmitting a specific series of pulses at a certain amplitude and period Amplitude : at least 3dB larger
Much larger than that of a regular data transmis-sion
Reliability of pulsing mechanism? Proved by “A Power Control MAC Protocol for
Ad Hoc Networks”.
15 / 31
Pulsing based solution (2)
WFP (Wait-for-First-Packet) pulses Used only for first packet reliability Short duration pulses Single radio Advertisement of incoming packets Negative ACK Simple energy detection
Different types of WFP Forced pulses Carrier sensing pulses Piggybacked pulses
16 / 31
Pulsing based solution (3)
A sink sends WFP pulses periodically Before it sends the first packet For a deterministic period
A sensor sends WFP pulses periodically After it receives WFP pulses Until it receives the first packet
WFP merits Prevents ACK implosion with small overhead Addresses the single or all packet lost problem Less energy consumption Robust to wireless errors or contentions
17 / 31
Pulsing based solution (4)
18 / 31
Pulsing based solution (5)
Implicit NACK
19 / 31
Pulsing based solution (6)
3 modes in delivery procedure for single/first packet 1. the advertisement that notifies the en-
suring single/first packet to all nodes with the forced WFP pulses
2. the delivery that sends the single/first packet through simple forwarding (for (ex)CSMA)
3. the recovery that sends NACKs using WFP pulses to request for the retransmis-sion of the single/first packet
20 / 31
Virtual infrastructure called core (1)
The Core An approximation of the minimum dominat-
ing set (MDS) of the network sub graph to which the reliable message delivery is de-sired. the set of local designated loss recovery
servers that help in the loss recovery process. Constructing the core during the course of
a single packet flood.
21 / 31
Virtual infrastructure called core (2)
Principle The retransmission
by neighbor is suf-ficient to recover the loss of the same packet of all neighbors around it
22 / 31
Virtual infrastructure called core (3)
Instantaneous Core Construction To approximate the MDS problem, we select a node at 3i
hop distance as a core node
Approximate number of hops from the sink to the sensor
23 / 31
Two-Phase Loss Recovery (1) Core recovery – first phase recovery
Out-of-sequence Packet Forwarding with A-map Propagation
Out-of-sequence : NACK implosion
Solve the above prob-lem : uses a scalable A-map (Available Map)
Overhead? The ratio of the
number of core nodes (10 – 30%)
A map request ra-tio (less than 1 %)
24 / 31
Two-Phase Loss Recovery (2) Non-core recovery – second phase re-
covery Starts only when a noncore node overhears an A-map from the core node indicat-ing that it has received all the packets in a message
25 / 31
Multiple Reliability Seman-tics (1)
26 / 31
Multiple Reliability Seman-tics (2)
Involving nodes employing a candidacy check before participating in the core construction algorithm
The candidacy check is where nodes, upon receiv-ing the first packet, deter-mine whether or not they belong in the subset G(s)
27 / 31
Evaluation (1)
For n2-based experiments 100 nodes in 650 m * 650 m square area Randomly deployed within that area Sink is located in center Transmission range of each node is 67 m Channel capacity is 1Mbps Each message : 100 packets (25 pkts/ sec) Size of packet : 1Kbyte
28 / 31
Evaluation (2)
Evaluation of single-packet delivery
29 / 31
Evaluation (3)
Evaluation of multiple-packet delivery
30 / 31
Discussion
Considerations for upstream? Network model
Sink and sensors static? There is exactly one sink coordinating the
sensors? Congestion control?
If congestions are appeared, how can GARUDA control them?
Loss recovery for noncore nodes How can we reduce snooping overheads?
31 / 31
Q & A