An Energy Efficient MAC Protocol for Wireless Sensor Networks “S-MAC”Wei Ye, John Heidemann, Deborah Estrin

Presentation: Deniz ÇokusluMay 2008

Outline Motivation S-MAC Experiments Results Conclusion Questions

Motivation WSN’s are battery operated, disposable(!) devices Energy consumption is an important issue Communication dominates most of the energy

consumption Medium Access is very critical

Motivation Medium Access Control (MAC)

Allows accessing transmission media Should avoid collisions

TDMA Reserves time slices to nodes (Time Synchronization)

CDMA Transmission is coded into signals (Extra computation)

FDMA Nodes are assigned to different frequencies (Complex

Xmitters) *CSMA

Media is shared (Collisions)

Motivation Contention-based protocols

MACA + RTS/CTS are used - Hidden & Exposed Terminal Problems

T-MAC + RTS/CTS, TimeOut - Early Sleeping

Preamble sampling Periodic sleeping, Long Preamble to wake up receiver

B-MAC Low Power Listening

PAMAS Use of Control Channel (RTS/CTS/Busy Tone)

Motivation IEEE802.11

RTS/CTS Fragmentation support CTS only reserves medium for the first fragment ACK of the first fragment reserves medium for the

second and so forth.

Overhearing problems Long delays because of the fragmentation faults Idle listening

Challenges *Energy efficiency *Scalability *Fairness Latency Throughput Bandwidth utilization

Sources of Energy Waste in WSNs Collisions

Causes retransmission of packets Overhearing

Receiving others’ packets Control packet overheads

Forms some part of the transmission Idle listening

Keeping the receiver powered on while the node is idle

Consumes %50 - %100 of energy required for receiving

S-MAC

Contributions of S-MAC Periodic listen/sleep

Nodes listens and sleep periodically Reduces idle listening

Collision and overhearing avoidance Uses RTS/CTS and carrier sense Puts a node into sleep if its neighbors are communicating Eliminates overhearing problem

Message passing Divides the message into fragments and send them in burst Reduces control overhead

Periodic Listen / Sleep Sleep some time

Turn off the radio Set the timer

Wake up Listen to see if there is a request for

communication

Requires periodic synch among neighbors Relative timestamps Small time slots

Periodic Listen / Sleep Synchronization

Neighbor nodes exchange their sleep/listen schedules

Two neighbor nodes may have different schedules

All nodes know their neighbors’ schedule When a node wants to send a message, it waits

until receiver wakes up If multiple nodes want to send, they contend the

media using RTS/CTS

Periodic Listen / Sleep Synchronization

Each node maintains a schedule table Choosing schedule

Listens if any neighbor broadcast a schedule If it receives a schedule, it uses as its own schedule

(follower node) and broadcasts it as a SYNC message If no schedule is received, choose the schedule

randomly and broadcasts it as a SYNC message (synchronizer node)

If the node receives a schedule after selecting its own, it uses both (multiple schedule)

Periodic Listen / Sleep Maintaining Synchronization

Periodically send SYNC packets SYNC packets are small, and includes sender’s next

sleep time Listen interval is divided into two:

Listening SYNC Listening RTS

If a node wants to send a SYNC or Data Wait until receiver wakes Start carrier sense and sense till random time If no xmission, send RTS and then data

Collision and Overhearing Avoidance Collision Avoidance

Carrier sense, RTS and CTS are used There is a “duration” field in all packets If a node receives a packet destined to another

node, it records the duration into Network Allocation Vector (NAV)

This value is decremented when the NAV timer is fired

If this value is greater than 0, it means that the medium is busy (virtual carrier sense)

Collision and Overhearing Avoidance Collision Avoidance

Physical carrier sense is performed at the physical layer

If both physical and virtual carrier sense indicates that medium is free than node uses the medium

Message Passing Deals with transfering long messages

efficiently Options

Long Messages: + Low control overhead - High cost of retransmitting in case of errors

Fragmented Messages: + Low cost of retransmission - High control overhead

Message Passing S-MAC Approach:

Small fragments of messages Cure to control overhead:

Send fragments burstly Use only one RTS/CTS packet for the whole message All fragments are confirmed by ACK messages

If any fragment is failed: Extend the reserved transmission time Retransmit the current fragment Hidden terminals learn the extension from the ACK packets Extensions are limited to a pre-defined number

S-MAC Trade-offs Delays inherent to multi-hop contention

based network protocols (i.e. IEEE802.11) Carrier sense delay Backoff delay Transmission delay Propagation delay Processing delay Queueing delay

Additional delay of S-MAC Sleep delay

S-MAC Trade-offs Avg sleep delay

A cycle consisting of sleep and listen is called a frame

Tframe = Tlisten + Tsleep

Ds = Tframe / 2

Energy saving

Es =Tsleep / Tframe

Sleep SleepListen

Frame

Implementation Rene Motes are used

Transmission rate: 19,2 Kbps Energy consumption:

Receiving/Idle: 4,5 mA Transmitting: 12 mA Sleeping: 5 μA

Motes use TinyOS Standard Packets in TinyOS

Header: 6B Payload: 30B CRC: 2B Total: 38B

Implementation Additional packet type: Control Packets

(RTS, CTS, ACK) Header: 6B Payload: NA (Benefit: 30 Bytes) CRC: 2B Total: 8B

Implementation 3 MAC Modules are implemented for

comparison Simplified IEEE802.11 DCF Message Passing with overhearing avoidance S-MAC

Implementation Simplified IEEE802.11

Carrier Sense Random duration

Backoff and retry Random backoff (unlike standard IEEE802.11

Exponential) RTS/CTS/DATA/ACK packet exchange Fragmentation support (Burst mode) Nodes are either in listen/receive or transmitting

mode

Implementation Message Passing with overhearing

avoidance Reduces control overhead Eliminates overhearing No periodic sleep, therefore no additional delay to

IEEE802.11 Node goes to sleep if its neighbors are

communicating

Implementation S-MAC

Listen time: 300 mSec Sleep time: 1000 mSec Schedule update (SYNC): 10 frames (13 sec)

Experiments Used Scenario:

Messages:A C DB C E

Energy consumption is measured Inter arrival period of messages varies between 1s – 10s

Ex. If period is 5s then a message is generated in every 5 sec by each source node

Experiments Used Scenario:

Each source generates 10 messages Each message consists of 10 fragments Total of 200 packets are exchanged in each

experiment In the higher traffic scenario inter-arrival: 1s, the

wireless media is fully utilized

Experiments Calculations of energy consumption:

Time to pass packets is measured Percentage of Listen – Sleep is measured Consumption is calculated by multiplying the time

with the required energy Receive/Listen: R = 13,5 mW Transmit: T = 24,75 mW Sleep: S = 15 μW

Total = (R * Tlisten) + (T * Ttransmit) + (S * Tsleep)

Experiments

Measured energy consumption in the source nodes

Experiments

Measured percentage of time that the source nodes in the sleep mode

Experiments

Measured energy consumption in the intermediate node

Conclusions S-MAC is an energy efficient mac protocol

compared to IEEE802.11 Trade-off between energy consumption and

latency can be easily tuned Experiments show that the aimed results are

satisfied

Acknowledgements This work is supported by:

NSF under grant ANI-9979457 as the SCOWR project (http://robotics.usc.edu/projects/scowr/)

DARPA under grant DABT63-99-1-0011 as the SCADDS project (http://www.isi.edu/scadds/) and under contract N66001-00-C-8066 as the SAMAN project (http://www.isi.edu/saman/) via the Space and Naval Warfare Systems Center San Diego

Questions ...