UESTC 4020: Wireless Sensor Networks

Transport Layer
Dr Hanaa Abumarshoud
[email protected]

1
REFERENCE TEXT

Chapter 8, Wireless Sensor Networks by Ian F Akyildizand

Mehmet C Vuran.

2
TRANSPORT LAYER OVERVIEW
▪The success and efficiency of WSNs directly depend on reliable communication between the
sensor nodes and the sink → a reliable transport mechanism is needed.
▪The main functions of a transport layer protocol for WSNs are:
1) Congestion control:
• Packet losses can happen due to congestion → reduced reliability
• Congestion control increases network reliability + helps conserve scarce sensor resources.
2) Reliable transport:
• Ensure that data sent by the sensor node reaches the sink, and vice versa.
3) (De)multiplexing:
• Bridge the application and network layers by using suitable multiplexing and demultiplexing to serve
multiple applications through the same WSN.
TRANSPORT LAYER OVERVIEW
▪There are several transport layer solutions developed for conventional wireless networks.
▪Many of these solutions are not suitable to accommodate the unique characteristics of WSNs.
Why?
1) Strict end-to-end reliability, which is based on acknowledgments and end-to-end
retransmissions, imposes significant overhead for implementation in WSNs.
2) The inherent correlation in the data flows generated by the sensor nodes makes these
mechanisms significantly energy inefficient.
3) Considerable memory capacity is needed to buffer transmitted packets until they are
ACKed by the receiver, but sensor nodes have limited buffering space and processing
capabilities.
 End-to-End Measures

Application-Dependent Operation

CHALLENGES
FOR TRANSPORT Energy Consumption

LAYER
Constrained Routing/Addressing

Biased Implementation
CHALLENGES FOR TRANSPORT LAYER:
END-TO-END MEASURES
Conventional transport layer solutions:
▪Provide end-to-end and point-to-point reliability and congestion control.
▪Packet losses and congestion mitigation are performed through communication between a source
and a destination without any involvement from the intermediate parties.
▪The transport control mechanisms reside only on the source and destination.
▪Each flow is considered independently to provide a point-to-point communication solution.
CHALLENGES FOR TRANSPORT LAYER:
END-TO-END MEASURES
The problem:
Collective information from a group of sensors is much more important than the individual
information from each sensor node → conventional end-to- end, point-to-point transport layer
techniques may lead to waste of scarce wireless sensor resources.
The solution:
▪Local measures for reliability and congestion control are usually exploited in WSNs to improve
the energy efficiency of the transport layer protocols.
▪Reliability of the collective information from a group of sensors is controlled instead of the
reliability of information from each individual sensor node.
CHALLENGES FOR TRANSPORT LAYER:
APPLICATION-DEPENDENT OPERATION
The problem:
WSNs are typically deployed with a specific sensing application objective and there is no one-
fits all measure.
Examples:
▪WSN for monitoring of a specific phenomenon → reliability is the most important metric.
▪WSN for fire event detection → timeliness is crucial.
The solution:
The design of the transport layer protocols should be tailored to the application and take into
account the design requirements and constraints of the specific deployment scenario.
CHALLENGES FOR TRANSPORT LAYER:
ENERGY CONSUMPTION
The problem:
Energy efficiency is the most important concern in the design of WSNs.
Typical protocols designed for conventional networking for usually require significant energy
consumption in a multi-hop network →not suitable for WSNs.
The solution:
▪Transport layer protocols should be energy aware, i.e., the error and congestion control
objectives must be achieved with the minimum possible energy expenditure.
▪Transport layer protocols can be designed such that the reliability level can be traded off for
decreased energy consumption through local reliability measures.
▪If reliability level at the sink are higher than the required level, the source nodes can conserve
energy by reducing the amount of information sent out or temporarily powering down.
CHALLENGES FOR TRANSPORT LAYER:
BIASED IMPLEMENTATION
The problem:
▪WSNs have a large number of resource-constrained sensor nodes that are connected to a single
resource-rich sink.
▪Sophisticated algorithms cannot run locally at the sensor nodes due to limited processing power
and memory capacity.
▪Traffic exhibits significantly different characteristics depending on the flow direction; flow in the
sensors-to-sink direction may require timely delivery with loss-tolerant operation, while the sink-to-
sensors direction usually requires a high delivery ratio.
The solution:
▪Transport layer protocols should be designed such that most of the functionalities are performed at
the sink with minimum functionalities required at the sensor nodes.
▪The protocols should be designed also by considering the biases in traffic flow direction.
CHALLENGES FOR TRANSPORT LAYER:
CONSTRAINED ROUTING/ADDRESSING
The problem:
▪Conventional transport layer solutions rely on end-to-end global addressing.
▪However, wireless sensor nodes may not be assigned unique addresses.
The solution:
Transport layer protocols should not assume end-to-end global addressing and utilise
attribute-based naming or data-centric routing instead.
CHALLENGES FOR TRANSPORT LAYER
Several transport layer protocols have been developed for WSNs:
▪Reliable multi-segment transport (RMST)
▪Pump slowly, fetch quickly (PSFQ)
▪Congestion detection and avoidance (CODA)
▪Event-to-sink reliability (ESRT)
▪GARUDA
▪Real-time and reliable transport (RT-2)
RMST: RELIABLE MULTI-SEGMENT TRANSPORT
▪The RMST protocol is one of the first transport layer protocols developed for WSNs.
▪The main goal of RMST is to provide end-to-end reliability.
▪RMST provides two of the three functionalities required for a transport layer protocol:
1) Reliable transport:
✓provides mechanisms to handle errors throughout the routes in the network.
✓utilizes in-network caching and provides guaranteed delivery of the data packets.
2) Multiplexing/demultiplexing:
✓Multiplexing at the sensor nodes and demultiplexing at the sink.
RMST
▪RMST is built on top of the directed diffusion routing protocol and uses some of its
functionalities.
▪Hence, an implicit assumption is that the packets of a flow follow the same path unless there is
a node failure.
▪In case of node failures, directed diffusion is assumed to reroute packets.

Directed diffusion routing

[network layer lecture]
RMST
RMST has two modes of operation:
1) Non-caching mode:
▪Similar to conventional transport layer protocols; only the source and destination play a role in
providing reliability.
▪Packet losses are detected at the sink and requested from the source node in an end-to-end fashion
through a NACK packet.
▪It requires no involvement from the intermediate nodes in the multi-hop network→ no additional
processing, storage, or energy consumption at the intermediate nodes.
2) Caching mode:
▪The intermediate nodes on the reinforced path cache the transmitted packets to decrease the
overhead in end-to-end retransmissions.
RMST
How does RMST provide reliability?
▪Each packet of a flow is labelled by a unique sequence number.
▪If there is a hole in the sequence numbers
→ a packet error is detected.
→ nodes request retransmission by sending a NACK packet toward the reverse route from the sink
to the sensor.
RMST
Example: non-caching mode
▪A sensor node is trying to transmit a series of packets to the sink
through a multi-hop route.
▪The sequence number of the last received packet at the sink is 3.
▪Packet 4 is lost before reaching the sink.
▪The sink can recognize the packet loss after it receives packet 5.
▪Then, the sink transmits a NACK packet back to the source node.
▪The lost packet is then retransmitted to the sink through the multi-
hop route.
RMST
Example: caching mode
▪Certain sensor nodes are assigned as caching nodes on the
reinforced path from the sensor node to the sink.
▪Loss packet detection is performed at the sink & caching nodes.
▪The caching node detects the loss of packet 3 after receiving
packet 4 and transmits a NACK packet back to the source node.
▪The first caching node with the missing packet on the reverse path
replies and the packets are transmitted to the sink in order.
▪If the packet cannot be found in one of the caching nodes, the
NACK packet is propagated until it reaches the source node.
RMST:
QUALITATIVE EVALUATION
▪RMST provides guaranteed delivery for each flow in the WSN & treats each flow separately:
→useful for applications where individual node information is important.
→may lead to overutilization of the resources in WSNs [most applications related to event
detection/tracking may not require 100% reliability since the individual data flows are correlated
and loss tolerant].
▪RMST follows a packet-by-packet reliability notion:
→a large amount of information may flow inside the network.
→this leads to congestion and associated packet drops, which is not addressed in RMST.
RMST:
QUALITATIVE EVALUATION
Caching vs non-caching:
▪Caching mode creates reliable segments between two consecutive caching nodes and the
retransmissions are performed inside these segments instead of through the end-to-end route.
→the cost associated with end-to-end retransmissions is minimized.
▪Caching requires additional processing and memory at the caching nodes
→this may increase the overall complexity and energy consumption of the network.
PSFQ: PUMP SLOWLY, FETCH QUICKLY
▪Contrary to many transport layer approaches that focus on the sensors-to-sink path, the PSFQ
protocol has been developed to address the path from sink to sensors.
Why?
▪The reverse path is generally used for network management tasking and re-tasking of the sensor
nodes → reliability is of major concern.
▪While the sensors-to-sink path may tolerate information loss due to the highly correlated data
generated by sensors, the reverse path requires one-to-one communication support to reliably
disseminate control information that is sent by the sink to each sensor.
PSFQ
PSFQ protocol provides three main functions:
1) Pump operation: PSFQ employs a slow pump mechanism that is based on slowly injecting
packets into the network. In this case, each node on the route to the destinations waits for a
specific amount of time before relaying the messages.
2) Fetch operation: In case of packet errors, each node performs aggressive hop-by-hop
recovery to fetch the lost packets from neighbour nodes.
3) Status reporting: PSFQ also provides a reporting functionality that creates closed-loop
communication between sensors and sink. Through this functionality, the sink can collect
information related to operation of the network.
Note: pump operation is the default strategy in PSFQ for information dissemination from sink to
sensors unless there are any errors.
PSFQ
PSFQ pump operation
▪Two timers, Tmin and Tmax, are used to schedule the transmission times of
nodes along a path from a sensor to sink.
▪A sensor broadcasts packets every Tmin to its immediate neighbours.
▪The neighbour nodes relay the packets after a random wait time, which is
selected between Tmin and Tmax.
▪A node has to wait at least Tmin between packet transmissions to allow nodes
to recover the missing packets.
▪Random delay allows a reduction in the number of redundant broadcasts of
the same packet by the neighbours.
▪If the packet is forwarded by one of the nodes, other neighbours suppress
their transmissions.
PSFQ
PSFQ fetch operation
▪The fetch operation results in local error recovery through persistent NACK
messages between two packet transmissions.
▪When node detects a packet loss, it broadcasts a NACK packet to its
immediate neighbours.
▪If the node does not receive any reply within a period of Tr, where Tr < Tmax,
it persistently continues to send NACK packets for every Tr.
▪If one of node’s neighbours has the packet in its cache, it sends the packet in
an interval between 0.25Tr and 0.5Tr.
PSFQ
PSFQ Status reporting
▪Report operation allows the sink to request feedback from the sensor nodes in a scalable manner.
▪Report operation is initiated by the sink by setting a report bit in the packet header.
▪This packet is sent through the network to the intended nodes.
▪Upon receiving the report request, a sensor node responds immediately by transmitting a report
message.
▪Each node along the path to the sink adds its status information to this report by piggybacking.
▪If a node in the upstream path does not receive the report response in Treport, it creates its own
report packet and sends it to the sink.
PSFQ:
QUALITATIVE EVALUATION
▪PSFQ employs a simple protocol operation that does not require an end-to-end reliability mechanism
→it scales well with network size.
▪PSFQ is rather a hop-by-hop reliability protocol than an end-to-end solution
→ end-to-end reliability cannot be guaranteed.
▪The pump slowly operation introduces an artificial delay in the network at each hop
→networks of large size may suffer from high latency.
▪Whenever a packet is lost, the receiver node stores the out-of-order packets until the lost packet(s)
are received
→ increased memory requirements.
CODA: CONGESTION DETECTION AND AVOIDANCE
▪The aim of CODA is to detect and avoid congestion in WSNs
▪Three main congestion scenarios are considered:
1) Congestion near the source nodes in WSNs applications in which the source nodes frequently
generate traffic.
2) Temporary congestion that occurs temporarily in hot spots where multiple flows are served.
3) Persistent hot spots congestion.
 CODA
▪A sensor node is trying to transmit a packet to the sink
through a congested area.
▪When the congestion is detected by one of the nodes inside
the congested area, the receiver broadcasts a suppression
message toward the source node in the reverse path.
▪The suppression message is used to inform the upstream
nodes about congestion.
▪Upon receiving the suppression, the upstream nodes decrease
their sending rates or drop packets to relieve the congestion
in the forward path.
▪The suppression message is rebroadcast until a non-congested
node receives the message by rerouting to avoid hot spots.
 CODA:
QUALITATIVE EVALUATION
▪CODA provide efficient congestion detection:
→the network performance is enhanced by avoiding congestion.
▪CODA only provides congestion avoidance and does not provide a reliable operation:
→end-to-end transport reliability is not guaranteed.
 ESRT: EVENT-TO-SINK RELIABLE TRANSPORT
▪ESRT protocol is based on the event-to-sink reliability notion, in contrast to conventional end-to-end
reliability.
▪ESRT aims to address both the reliability and congestion problems in WSN.
▪ESRT provides reliable event detection without any intermediate caching requirements.
 ESRT
▪The main characteristic of information delivery in WSNs is the data-centric communication
paradigm
→ collective information from multiple sensors regarding an event is much more important than the
individual information sent from each node.
▪The notion of flow (which is defined as the messages that are sent from a source to a destination),
no longer applies.
▪Instead, an event-to-sink information flow exists, which is the collection of “flows” from a group of
sensors associated with a single event.
 ESRT
▪The algorithms of ESRT run mainly on the sink, with minimal functionality required at resource-
constrained sensor nodes.
▪The data flows generated by the sensor nodes toward the sink is correlated due to spatial and
temporal correlation among the individual sensor readings.
▪The reliability is measured in terms of the number of data packets from all the sensor nodes
associated with an event.
➢Decision interval 𝜏: ESRT relies on the sink to measure the reliability in the WSN every 𝜏 time units.
➢Observed event reliability, 𝑟𝑖 : number of received data packets in decision interval 𝜏𝑖 at the sink.
➢Desired event reliability, 𝑅: is the number of data packets required for reliable event detection.
This is determined by the application.
 OTHER TRANSPORT PROTOCOL
➢GARUDA:
▪Incorporates an efficient pulsing-based solution for successful delivery of single packets.
▪Certain nodes in the network are selected to perform caching and manage loss recovery process.
➢(RT)2: Real-Time and Reliable Transport
▪(RT)2 has been developed to reliably and collaboratively transport event features from the sensor
field with minimum energy dissipation in a timely manner.