Institut für Telematik | Universität zu Lübeck

Wireless Sensor Networks

Chapter 5: MAC Protocols

Stefan Fischer
Dennis Pfisterer
Learning Goals 5-2

 Understand what is different about radio MACs and especially sensor

network MACs
 Know the most important MACs
 Know the differences in efficiency
Content 5-3

 Requirements for MACs in sensor networks

 Contention-based MACs
 CSMA/CA
 S-MAC
 T-MAC
 B-MAC
 WiseMAC
 Schedule-based MACs
 LMAC
 Bluetooth
 Zigbee
 Evaluation
Media Access (MAC) 5-4

 Controls access to shared medium

 Avoid mutual disturbances
 Part of ISO OSI data link layer (2)
 Important factor for energy consumption
 Radio is primary consumer
 Main goal: reduction of energy consumption
Energy Consumption in the Radio 5-5

 Energy Consumption in the Radio

 Send / receive = 1 ... 2
 Receive / listen = 1 ... 2
 Listen / sleep >> 100
 Mode switches can take a while
- 10us – 2ms
 Implications
 Sleep as much as possible (but then you are deaf)
 Switch the mode only seldomly
Waste of Energy 5-6

 Where is the potential for energy savings in existing protocols?

 Useless listening
 ... for incoming packets

 Collisions
 Messages get lost through signal interference of simultaneous
transmissions
 Receiver hears several senders simultaneously
 Sender cannot recognize collisions
- Not in the reception area of other senders
-  hidden-terminal problem

 Overhearing
 ... of messages destined to other receivers

 Protocol overhead
 Packet headers
 Additional messages
Tradeoffs 5-7

 In order to save energy, one has to accept compromises with other

properties
 Additional messaging delays, especially over multi-hop
 Throughput (payload per time)
 Fairness (nodes are not discriminated)
Traffic Patterns 5-8

 Further saving potential through observation/usage of typical pattern

in the message flow
 Message flow
 Convergecast: all sensors send to the sink (with aggregation)
 Broadcast: sink sends to all nodes
 Local interactions: neighbors exchange messages among
each other
 Geo Routing: send to a certain position
 Message generation
 Constant data streams
- Regular sensor readings
 Bursts
- Silence for long periods
- Events trigger activity in nearby nodes at the same time
 Application-specific properties
 Cross-Layer!
MAC Protocols 5-9

 Two classes:
 Contention
 Random access to the medium
 Collision avoidance as far as possible
- Listening on the medium (CSMA), random waiting, reservations
 When collision: re-transmission
 Schedule
 Exclusive assignment of the medium to the nodes
- Time slots (TDMA)
 Node is only active during ist time slot, sleeps otherwise
 No collisions, no unnecessary listening or overhearing
 BUT:
- Need for time sync
- Adaptivity?
CSMA/CA 5-10

 CSMA: listen on channel, send when free

 Problem: „hidden terminals“
 CA: reservation of the channel
 RTS: Request To Send with announcing the necessary
sending time T
 CTS: Clear To Send: acknowledgement with T
 DATA: after receiving CTS
 ACK: after receiving DATA
 All nodes which hear RTS/CTS delay messages by T

T
CSMA/CA: Collisions 5-11

 Collisions are possible anyway

 Which ones?
 Recognition through missing ACK
 Re-send after random delay
 Delay strategies
 Randomly chosen, equally distributed delay from
a given time window which is split in slots
 Problem: with many nodes you have many collisions
1 2 3 4 5 6
- Especially due to message bursts!
 Binary Exponential Backoff (BEB)
 Double the time window after each new
collision
 Adapt to number of nodes
 Problem: very long delays may occur
Sift 5-12

 Anternative for BEB

 Fixed time window, but slot selction is not
equally distributed
 Only a few nodes choose an early slot
 Probability for collision is smaller for early slots

(1   ) CW r
pr  
1 CW
CSMA/CA for WSN 5-13

 Useless listening
 Switch off radio during T
 Problem: other RTS/CTS may be missed
- Additional collisions
- Throughput may be up to 25% less
 Protocol Overhead
 RTS/CTS/ACK for every packet!
- Inefficient with small packets
 Single reservation for several packets
- Fairness vs. energy!
S-MAC 5-14

 CSMA/CA with Duty Cycling

 Brief activity phase followed by long sleeping phase
 Nodes are synchronized among each other
 RTS/CTS only during active phase
 Data will be transferred before and into the sleeping phase
 SYNCH/RTS/CTS
 CSMA phases with random
delay Active period

 If unsuccessful try in the next Wakeup period

active phase
Sleep period

For SYNCH For RTS For CTS

S-MAC: Synchronization 5-15

 Nodes send schedule packets during SYNCH

 New node
 Listen and accept schedule
 Otherwise use own schedule
 Problem: neighbors with diferrent schedules
 „islands“ have to be integrated
 Use schedules of all neighbors

Schedule 1 Schedule 2
S-MAC: Latency 5-16

 Expected delay per hop = half the length of the

sleeping phase
 Problem with long paths!
 Improvement:
 Path: 1 -> 2 -> 3
 Node 3: hears CTS, after T additional active phase
 Node 2: send an RTS directly after ACK 1 2 3

T
T-MAC 5-17

 S-MAC: fixed length of activity phase

 Unused capacity when not much traffic
 Too short with much traffic
 T-MAC: adaptive length of the active phase
 Early switch to sleeping phase if nothing happened for a certain time TA
 „nothing“
- No activity on the channel and
- No activity at neighbors (we know this through RTS/CTS)
T-MAC: Early Sleeping 5-18

 Problem: D starts sleeping too early

 C has data for D, but cannot send
RTS to D because medium is blocked
by CTS
 Solution: special RTS for the future
 When receiving CTS, node C sends
FRTS to D
 contains T from CTS
 additional dummy packet DS
- Why?
Preambles 5-19

 Problem: both S-MAC and T-MAC need synchronization

 Several schedules per node ...
 Different approach
 Unsynchronized, very short activity phases
 Sender has to send long preamble before data
- How long?
 If node hears preamble during active phase, he remains awake in order
to receive data
 Moves the load to the sender

Preamble Data
Preambles: Optimizations 5-20

 Problem: full sending of preamble

though station is already awake Preamble Data
 Waste of energy
 repeated sending of data instead
of preamble
 Repeated sending of wakeup packets
instead of preamble
Data Data Data
 Contains point in time when data
transmission starts

W WWWWWW Data
B-MAC 5-21

 Is based on simple preambles

 Philosophy: minimum functionality with simple interface (TinyOS)
 Length of sleeping phase is adjustable
 Length of preamble is adjustable
 Fixed-length active phase with preamble detection
 CSMA with simple random backoff
 Optional ACKs
 Additional functionality has to be implemented separately
 RTS/CTS, ...
 Maximum flexibility
 Standard MAC in TinyOS
WiseMAC 5-22

 Preambles and synchronization (!)

 First, a long preamble
 ACK contains schedule
 Now, the sender roughly knows, when the receiver will wake up the next
time
 Shorter preamble
 Additional specialties
 „More!“ Bit in data packet for bursts
 Repeat data instead of long preambles

Preamble Data ACK P Data ACK

MAC with Schedules 5-23

 Division of time in slots

 Every node gets some slots during which it is allowed to send
 Many variants in assigning slots
 centralized vs. distributed
 Static vs. usage-oriented
LMAC 5-24

 32 time slots per frame

 Every node gets one slot
 In ist slot, node can send messages to neighbors
 Neighbors hear TC, sync
 Only neighbors, which are addressed, also hear the data
 No ACK, no retransmission when failure occurs

Slot 1 Slot 2 Slot 3 ... Slot N Slot 1 Slot 2 Slot 3 ...

TC Data

Snd Rcv Slots ...

011011101000000...0000
LMAC: Slot Assignment 5-25

 TC.Slots contains a bit vector of all used slots

 As seen by the sender
 New node
 Listens to neighbors for one frame (TC.Slots) and then choises randomly
a free slot
 Mobility may lead to collisions (double slot usage)
 Colliding stations randomly choose a new slot

1011
0011
3
4

1 2
1010
LMAC: Restrictions 5-26

 Density of the network is restrictred by the number of slots

 Neighbors must not use the same slot
 Maximum bandwidth per node is limited
 Channel bandwidth / number of slots
 Bad when communication volume of nodes varies a lot
Bluetooth 5-27

 Bit transmission
 79 channels
 Pseudo-random frequency hopping S
S
 Bandwidth ~ 1 mbps
M
- Energy efficient with high data volume S S
 MAC based on scheduling
B
 Piconet: 1 master + max. 7 slaves S
- Master synchronizes slaves and announces M
S
hopping pattern S
- Every slave has a time slot during which S
it can communicate with the master
 Scatternets: overlapping piconets
- Bridge: common node
IEEE 802.15.4 („Zigbee“) 5-28

 Bit transmission
 20 – 250 kbps in different frequency ranges (868 kHz, 926 kHz, 2.4 GHz)
 Usage of multiple channels possible
 Roles of nodes
 PAN Coordinator (always active)
 Coordinator (always active!) D
D
 Device
P
 Node types
D D D
 Full Function Device
D
 Reduced Function Device D
- just „Device“ role D C
 Network topologies D
P
 Star D
 Peer-to-Peer
Other Approaches 5-29

 There is a plethora of further proposals for WSN MAC protocols

 We looked at the most relevant ones
 Other assumptions concerning bit transmission
 So far: all stations use the smae channel/frequency
 Chossable channel (as in Bluetooth)
 Several channels at the same time
 Several radions per node
- Wakeup Radio
 Directed antennae instead of omni-directional ones
 ...
The best MAC? 5-30

 Is there a „best“ MAC protocol?

 Resource need
 Energy efficiency
 Throughput
 Loss of data
 scenarios
 Point-to-point
 Local interactions
 Protocols
 802.11: Standard CSMA/CA
 S-MAC
 T-MAC
 LPL: preambles (B-MAC)
 LMAC
Resources 5-31
Energy when inactive 5-32
Throughput: 3x3 Grid 5-33
Energy: Point-to-Point 5-34
 5-35
Energy: Local Interaction
Summary 5-36

 „winner“ wrt. energy

 Low data volume: T-MAC
- Through adaptivity
 High data volume: LMAC
- No collisions
 Further observations
 802.11 is very competitive with high data volume
 There is no „best“ MAC, depends on the application