CCR: A Novel MAC Scheme with ConstantTime Contention Resolution for WLAN

1
CCR A Novel MAC Scheme with Constant-Time
Contention Resolution for WLAN

• Zakhia G. Abichar and J. Morris Chang
• Computer Systems Laboratory
• Dept. of Electrical and Computer Engineering
• Iowa State University
• Ames, IA 50011
• December 10, 2004

2
Motivation

• WLAN ubiquitously employed
• DCF Basic access scheme
• Used for PCF and the proposed HCF
• Based on the CSMA/CA algorithm
• Contention Window (CW) management based on the
  binary exponential backoff (BEB) algorithm
• High collision rate for a large network size
• CCR addresses the issue of high-collision rate in
  DCF

3
Objective

• Propose a scheme that
• Is distributed
• Exhibits a low collision rate (even at large
  network sizes)
• Resolves contention in a constant time

4
Outline

• Overview on Wireless MAC Protocols
• The IEEE 802.11 Standard
• The CCR Scheme
• Mathematical Analysis
• Performance Evaluation

5
Overview

• Wireless networks
• Shared medium
• Basic MAC scheme
• The most delicate component of wireless network
  architecture

6
Overview

• Research in wireless MAC schemes started in the
  1970s
• Pure ALOHA
• A station transmits whenever a packet is ready
• No carrier-sense
• Medium utilization 18
• Slotted ALOHA
• Time is divided into discrete slots
• Transmissions occur only at slot boundary
• Medium utilization 36
• CSMA protocols MACA, CSMA/CA
• Carrier-sense capability
• Avoid colliding with ongoing transmissions
• Use of RTS/CTS

7
Overview

• Wireless MAC schemes (contd)
• Fast Collision Resolution (FCR)
• Allowing bursts of packets
• Issue of fairness
• Applying SCFQ (Self-Clocked Fair Queuing)
  algorithm to mitigate the fairness metric
• Blackburst
• Adopted in HIPERLAN
• STAs jam the medium to indicate their delayed
  time
• Low throughput for large network size

8
Overview

• CCR is based on the Binary Countdown Mechanism
• Contention runs for d slots
• d is the number of bits in stations address
• A station jams for a 1 and senses for a 0
• Station with higher address always wins
• Based on a mathematical analysis CCR determines
• A proper number of contention slots
• Optimal probability of choosing 1 in a given
  time slot

9
Overview

• Binary tree schemes
• Close to the binary countdown mechanism
• Allow collisions to occur and then split
  colliding stations into two groups
• Not a good idea for wireless networks since
  collisions cannot be detected
• Difficult to apply in wireless networks
• Need multi-channel setting and feedback
• CCR is free of the above limitations

10
The IEEE 802.11 Standard
CSMA/CA algorithm
DCF Basic access mode
Alternates with DCF
PCF Contention-free access mode
Alternates with DCF
HCF (under study, not yet standardized)

• All of the above schemes rely on the efficiency
  of the CSMA/CA algorithm

11
CSMA/CA Algorithm

• Two steps to access the medium
• Wait for inter-frame space (IFS) to expire
• Decrease back-off timer

12
CSMA/CA Algorithm

• Reasonable performance for best-effort packets
• Large decline in throughput for a large network
  size
• Most wasted time slots attributed to backoff
  window slots and collisions

Average number of collisions per transmission
E Nc
Average number of back-off slots per transmission
E Bc
Value of aSlotTime
ts
13
The CCR Scheme
14
CCR Scheme The Rationale

• As shown in the previous slide, most of the
  wasted time in the DCF scheme is due to
• Collisions (dominant factor)
• Backoff slots
• The main goal of CCR is to reduce the number of
  collisions

15
CCR Scheme

• Allows a channel access after a constant number
  of time slots
• Exponentially reduces the number of contending
  stations every contention slot

t 2
t 1
16
CCR Scheme Operation

• Each station chooses a random value for its
  try-bit (stations divided into active group and
  passive group)
• Active group stations jam the medium if they want
  to transmit while passive group stations must
  sense the medium
• If the medium has been jammed, passive group
  retires from the contention. Otherwise, passive
  group remains
• Remaining stations refresh their try-bit and new
  groups are formed
• Contention time is of logarithmic complexity

Passive Group
Active Group
17
CCR Access Example
Low probability for p
Higher probability for p
Choose a value of 1 with probability p Run
CCR for k slots
18
CCR Scheme In Practice

• Choosing a try-bit of one with probability ½ is
  not the best choice
• Better have a low value at the early stages of
  the contention, then a higher value later on

• Contending stations need to know the number of
  time slots to run CCR
• Run the scheme for the same number of slots for
  10, 50, 100 stations

19
Mathematical Analysis
20
Mathematical Analysis

• Objective of the analysis, threefold
• For how many contention slots to run CCR?
• Choosing a value of one for the try-bit with what
  probability?
• Estimate the throughput of CCR

21
Try-bit equals to one with proba. p

• Probability that i stations move to the next
  contention slot

• Expected value of i

22
Number of Contention Slots

• Number of contending stations converges to one
  after 3 time slots
• Perfect knowledge assumed (figure 4)
• Simulation studies 6 slots are used with
  probabilities
• p 0.07, 0.2, 0.25, 0.33, 0.4, 0.5
• Note Values of p remain unchanged when the
  number of stations varies

23
Throughput Estimation

• Transmission interval
• Proba. of a transmission (Markov Chain)
• Proba. any STA transmitsps n . t . (1 t)n-1
• Proba. of a collision

k number of slots Note collision is the
opposite of successful transmission plus the
nothing to transmit.
24
Throughput Estimation

• In a transmission interval, there could be
  several collisions and one successful
  transmission
• Mean length of an interval
• Where p Nc j (pc)j . ps
• Throughput is given by
• where Ts is the transmission time of a packet.

j number of collisions
25
Throughput Estimation

• From the model, the throughput of CCR reaches 95

26
Performance Evaluation
27
Throughput at 2 Mbps

• Throughput of CCR reaches 92, 91, 90
• Throughput of DCF reaches82, 66, 58

28
Throughput at 11 Mbps

• Throughput of CCR reaches 87, 86, 85
• Throughput of DCF reaches79, 64, 57

29
Throughput Observation

• At low packet sizes and small network size (10
  STAs)
• DCF outperforms CCR
• Average access time of CCR is larger than DCFs
• The lower collision rate of CCR does not
  materialize
• Packet sizes are small
• Observation holds for packets
• Smaller than 50 bytes at 2 Mbps, (300 mirco-s)
• Smaller than 475 bytes at 11 Mbps, (300 micro-s)

30
Collision Rates

• Collision rate of CCR ranges from 4.37 to 6.47
• Collision rate of DCF ranges from 16 to 40
• Main reason why CCR outperforms DCF
• The high collision rate of DCF impedes other
  metrics, i.e. throughput, delay, fairness

31
Delay Measurement

• Worst-case delay of CCR is smaller
• Delay distribution of DCF is much fluctuating
  (caused by the higher collision rate)

32
Fairness Metric

• Longer time range for DCF to exhibit fairness
• In the short-run, colliding stations are penalized

33
Conclusions

• CCR achieves the goals set in the objective by
  allowing
• A higher throughput in the vast majority of cases
• A low collision rate even at large network sizes
• A smooth delay distribution
• A fairness allocation of the network bandwidth

34
Questions?

• Thank You