1. Topics in Distributed Wireless Medium Access Control
Nitin Vaidya
© 2010

2. Outline Service differentiation Rate control Power control
Priority scheduling
Throughput fairness
Rate control
Power control
Impact of directional antennas

3. IEEE 802.11 Distributed Coordination Function (DCF)
Physical & virtual carrier sensing (RTS-CTS)
Contention window (cw) : Backoff chosen uniformly in [0,cw-1]
Exponential backoff after a packet loss
Contention window reset to CWmin on a success
Inter-frame spacing: SIFS & DIFS

4. IEEE DCF

5. Service Differentiation at MAC Layer
Priority scheduling
Throughput fairness

6. Priority Scheduling Priority assigned to packets
Goal: When two packets are competing for transmission, packet with higher priority should “win” access first
Let us consider two priority levels, high & low
Ideas can be extended for more levels

7. Priority Scheduling Two packets at the same host
Transmit low priority packets only if high priority queue empty
Low
priority
High
priority

8. Priority Scheduling Packets at the different hosts
How to ensure that high priority packets will be transmitted first?
Host A Host B
High
priority
Low
priority
Low
priority
High
priority

9. Priority Scheduling: Inter-frame Spacing Differentiation

10. Inter-frame Spacing Differentiation
IFS for new high priority packet : DIFS
IFS for new low priority packet: LIFS LIFS = DIFS + Bmax where Bmax is the largest backoff interval used by a high priority packet

11. Inter-frame Spacing Differentiation
High
priority
Low
backoff
data
DIFS Bmax
LIFS D

12. Issues
Low priority traffic waits for LIFS even when no competing high priority traffic from other hosts
Different hosts maysee different channel conditions
A host with high priority packet may observe channel busy, while another host may observe it idle
 A low priority packet may be transmitted when a competing high priority packet is pending

13. Priority Scheduling: Backoff Interval Differentiation
Different range of backoff intervals for packets of different priority
Example: High priority: [0,H]
Low priority: [H+1, L]

14. Priority Reversal May Occur: Different Packet Arrival Times
Different hosts may start “counting down” backoffs at different times
High priority

15. Priority Reversals May Occur: Collisions
Priority Low
High
freeze backoff
data X Y Z
denotes backoff

16. Priority Reversals May Occur: Collisions
Priority Low
High
freeze backoff
data X Y Z
Can we improve the situation by changing actions for low priority?
denotes backoff

17. Throughput Fairness

18. IEEE 802.11 DCF “Symmetric” protocol
All things being equal, different single-hop flows should achieve similar throughput
In general, performance may not be “fair”

19. Potential Causes of “Unfairness”
Short-lived flows:
Random selection of backoffs may cause short-term unfairness
Collisions & exponential backoff
Spatial variations
Number of flows per host
Channel variations
If host with poor channel transmits at high rat  packet losses
If it transmit with low rate  uses up more channel time

20. What is Fair ?
Due to channel variations, and differences in interference levels, it is not always reasonable to requires all flows to achieve same throughput
Alternative definitions of fairness

21. Throughput Fairness
Only one host may transmit to C reliably at any time
11 Mbps
2 Mbps

22. How to achieve (approximate) these rates?
Adapt backoff to achieve this?

23. Service Differentiation: Throughput Fairness
To be discussed later

24. Rate Control

25. Rate Control Consider node A sending packets to node B
Choose rate based on channel conditions
Better channel  Higher rate
How to estimate channel conditions?
Implicit versus explicit feedback

26. Implicit Feedback Ack received
 Channel good enough for currently used transmission rate
Ack not received  Chosen rate too high

27. Rate Adaptation with Implicit Feedback
X = 3 , Y = 2

28. Rate Adaptation with Implicit Feedback
initial Xi = 3 , Y = 2
Xi adapted dynamically

29. Explicit Feedback
Receiver can send current SINR estimate along with an Ack
Transmitter can choose rate based on SINR estimate
Interaction with the NAV (virtual carrier sensing) mechanism

30. Power Control

31. Distributed Power Control
Transmitter i transmitting to node ri
Transmitter i need to achieve SINR

33. Iterative Algorithm
Converges if all transmissions are feasible above desired SINR threshold

34. Power Control with Interference Margin Dissemination
Parameter q
A transmitting data to B:
Interference margin at B = M
Busy tone D
power = q/M
data B A C E

35. Power Control with Interference Margin Dissemination
Parameter q
Interference margin at B = M
Busy tone D
power = q/M
data B A C
Transmit power < q / Pb
Received busy-tone power = Pb E

36. Analysis Interference by C at B = (q / Pb) gCB where Pb = (q/M) * gBC
= M * gBC / gCB
= M if gBC = gCB
= Interference margin

37. Issues gBC and gCB may not be equal: different channels
Multiple interferers

38. Diversity Multi-channel Multiple antennas
Directional (beamforming) antennas

39. Directional Antennas

40. Issues Impact on virtual carrier sensing
Even if C receives CTS from B, may be OK to transmit to D directionally
Potential solution: directional NAV

41. Issues Impact on physical carrier sensing
E cannot sense A’s directional transmission to B
E might collide at B

42. Issues Interaction with collision avoidance A transmitting data to B
Unaware that A is busy, E attempts to transmit to A, but no response is received
E assumes collision, and performs exponential backoff
Process may be repeated
While E is counting down a long backoff, A may finish transmitting to B, count down a short backoff, and again begins transmitting to B
 Short-term unfairness, or packet loss at B

43. Chapter Summary Service differentiation Rate control Power control
Priority
Fairness
Rate control
Power control
Directional antennas