CSE 592 MAC 1

1
Medium Access ControlHimanshu Gupta
2
Network Protocol Layers
Applications
Transport Layer (TCP variations)
Determines routes (multi-hop) to a destination.
Network Layer (Routing/Addressing)
Resolves interference conflicts, and schedules
link transmissions
Link Layer (Medium Access Control)
Physical Layer (Radio)
Handles transmission of raw bits
3
Network Graph Abstraction

• For Routing and Topology Control, we abstracted
  the network as a graph
• Edge (u,v) signified that u and v can communicate
  directly with each other.
• How does a node communicate with other, in
  presence of possibly other simultaneous
  transmissions?

4
Wireless Interference

• When a sender transmits, neighboring nodes hear
  the signal.
• Consider a receiver U
• If only one sender in vicinity, then no
  problem.
• If multiple senders in the vicinity
• Either, U retrieves none of the signals
  (collision).
• Or, U is able to retrieve the most dominant
  signal.
• We assume the first case (collision) for now.

5
Wired Vs.Wireless

• Carrier-sense multiple access (CSMA) listen
  before talk.

B
Ethernet LAN
Wireless LAN
C
A
A
B
C

• If both A and C sense the channel to be idle at
  the same time, they send at the same time.
• Ethernet Collision can be detected at sender.
• Wireless Half-duplex radios and limited
  transmission range. Collision can NOT be detected
  at sender.

6
Hidden Terminal Problem
C
B
A

• A and C cannot hear each other.
• When A transmits to B, the node C cannot detect
  the transmission.
• If C transmits as well, packets collide at B.
• A and C are hidden from each other.

7
Hidden Terminal Problem

• Cannot happen in wired Ethernet. Peculiar to
  wireless.
• Need to sense carrier at receiver, not sender
• Solution Do virtual carrier sensing. Ask
  receiver whether it can hear anything. If it
  does, behave as if channel busy.
• This idea is the core of 802.11 (Wi-Fi) wireless
  LAN protocol.

8
Exposed Terminal Problem
B
A
E
D

• In principle, A-gtB and D-gtE transmissions can go
  in parallel without collisions.
• But D hears As transmission, and waits.
• D is exposed to As transmission.

9
Virtual Carrier Sensing(Used in IEEE 802.11
standard)

• When A wants to send a packet to B
• A sends a Request-to-Send (RTS) to B.
• B responds by sending Clear-to-Send (CTS).
• Only nodes that hear CTS keep quiet for a given
  period (part of RTS and CTS packets).
• Thus, C keeps quiet, while D can transmit (to E).
• Solves both hidden and exposed node problems.

B
A
C
D
E
..
Example A sends to B
10
802.11 Standard More Details.

• If ACKs are required (from B to A after data
  transmission), then nodes that hear RTS should
  also remain quiet (exposed terminal problem
  resurfaces).
• Some delays (SIFS, DIFS) are introduced to avoid
  RTS/CTS collisions.
• Exponential back-off mechanism to resolve
  contentions.

11
Many MAC layer Issues

• Multi-channel MAC protocols
• With or without a control channel
• Directional antennas
• Improves capacity by lowering interference.
• How to design MAC protocols?
• Power-control MAC protocols
• Allows change in transmission range per-packet
• Fairness
• Whether all nodes are getting an equitable share
  of bandwidth. May require exchange of control
  information (or a centralized controller).

12
Multi-Channel MAC Protocols-I

• Two Radios and a Control Channel
• First radio is used (in a dedicated control
  channel) to agree upon the channel for data
  transmission.
• Second radio is used for data transmission.
• One Radio, Split phase
• Time divided into control and data phases.
  Control phase is done in the control channel.
• One Radio, Fixed Channel Hopping
• Nodes hop through the channels in a fixed pattern
• If A and B need to transmit, they do RTS/CTS,
  stop hopping (other nodes continue hopping),
  transmit data, and then continue hopping.
• One Radio, Involved Channel Hopping SSH Talk

13
Multi-Channel MAC Protocols-II

• Receiver Directed
• Each node when idle stays on a quiescent channel.
  Sender moves to receivers quiescent channel to
  negotiate the data channel.
• Use of Busy Tones

14
Directional Antenna MAC

• Omni-directional Antenna Signal propagates in
  all directions.
• Directional Antenna Signal is directed in a
  cone towards a direction of choice
• Helps in reducing interference, increasing range.
• Conceptually, similar to multi-channel MAC
  protocols.

15
TDMA Scheduling

• Back to network graph model. ?
• One way to schedule transmissions is to
• Divide time into slots
• In each slot, schedule a set of links (i.e.,
  edges of the network graph) that dont interfere.
• No carrier sensing/protocol messages here, but
  need tight time synchronization.
• Model interference
• Design algorithms to partition links into slots.

16
Modeling Interference

• Pairwise Interference
• Represented by a conflict graph over links, where
  there is edge (e1, e2) in the conflict graph if
  links e1 and e2 cannot be in operation
  simultaneously.
• Simplification of reality, but models many cases
  viz., protocol-based, distance-based models.
• Physical Interference
• Reception is successful iff SINR (signal to noise
  ratio) is more than a certain constant.
• SINR (Intended signal)
• (Noise sum of other signals )

17
TDMA for Pairwise Interference

• Given a network and its conflict graph (over its
  links)
• Partition the links into minimum number of time
  slots, such that in each slot no two links
  interfere.
• Same as partition the vertices of the conflict
  graph (links) into minimum number of independent
  sets.
• Assumption One copy of each link, i.e., uniform
  traffic.
• For general conflict graphs, NP-complete.
• We look at special/practical cases.

18
TDMA in Primary Interference

• Primary Interference
• Simplest case of pairwise interference
• Any two links incident at a common node interfere
• Rationale Each node has one radio (cant receive
  and transmit at the same time)
• TDMA in Primary Interference
• Partition the links into minimum number of
  matchings chromatic index of a graph.
• NP-complete.
• Greedy scheduling gives 2-approximate Why?

19
TDMA for Secondary Interference

• Secondary Interference
• Any two links connected by a link interfere.
• Also called 1-hop or Protocol.
• NP-complete
• For unit-disk network graphs, we can get a
  constant-factor approximation.
• Greedy Scheduling
• Iterate through links, putting each link in the
  first possible (or new) slot without creating
  interference with already placed links

20
Approximation Ratio of Greedy

• Greedy Scheduling
• Iterate through links, putting each link in the
  first possible (or new) slot without creating
  interference with already placed links
• Let X be the number of slots created.
• A link was put in the X-th slot, because it
  interfered with some link in each of the previous
  (X 1) slots.
• Claim At most 8 of these (X 1) links can be
  put in one slot by any algorithm. Why?
• Thus, optimal needs at least (X 1)/8 slots.
  Thus, Greedy is 8-approximate.
• 8 can be improved to 6, if links are considered
  in a particular order.

21
At most 8 Non-Interfering Links
How many nodes can be placed in the 2-cricle
region that are r-distance apart from each other?
Easy 55 10. Hard to prove 8 see
figure
r
r
r
The above shows that three can be at most 8
non-interfering links that interfere with a given
link.
22
Other Pairwise Interference Models

• K-hop Interference
• Any two links connected by a path of k-hops
  interfere.
• Distance Interference
• Any two links within a distance of d interfere
• Greedy Scheduling (and its proof) can be extended
  for above models.

23
TDMA for Physical Interference

• Signal received at a distance d from a sender
  operating at power P
• P/d?
• Reception at a receiver node R from a node u is
  successful iff the following equation holds (here
  dx denotes the distance of R from a node x, and N
  is the ambient noise)

• TDMA Scheduling Much more involved Presentation

24
Other TDMA Topics

• Multi-channel Multi-Radio
• Previous techniques can be extended.
• Joint Routing and Scheduling Marathe et al.
  SIGMETRICS 2005
• Joint Routing, Channel Assignment, and Scheduling
  in General Interference Ayyoub and Gupta, Tech
  Report 2008