TDM-based Coordination Function (TCF) in WLAN for High Throughput

Chaegwon Lim and Chong-Ho ChoiSchool of Electrical Engineering and Computer Science and ASRISeoul National University, Seoul, KoreaIEEE Global Telecommunications Conference (Globecom) 2004 (Acceptance rate: 37.7%)

Overview

Introduction Previous work Protocol Descriptions Performance Evaluation My Personal Comments Future work Conclusion

Introduction - 1

The demand for high throughput of IEEE 802.11 is increasing as most applications require a wide bandwidth.

However, the most popular protocol on IEEE 802.11 networks, Distributed Coordination Function (DCF), cannot meet the expectation due to its contention-based nature.

Introduction - 2

Collision resolution scheme in DCF, the Binomial Exponential Backoff (BEB), doubles the size of the contention window up to a maximum value (1024) when there is a collision.

A significant number of redundant idle slots are introduced due to successive collisions before a successful data transmission.

Therefore, the actual throughput is far below the theoretical one.

Introduction - 3

Factors considered in IEEE 802.11 networks’ performance analysis: The number of active stations. Bit Error Rate (BER) Contention window (CWin) size…

The number of active stations is directly related to the competition level in seizing the radio resource.

This germinates the idea of TCF.

Introduction - 4 Assumptions:

Propagation delay is negligible. all stations are within the radio transmission range no hidden node.

TCF uses information on the number of active stations explicitly to eliminate the contention period in DCF of IEEE 802.11.

It makes each station adjust the starting time of its radio transmission according to the number of active stations in order to avoid collisions.

Introduction - 5

It can be employed in both infrastructure and ad hoc modes and implemented distributively.

The overall throughput is improved as it is always higher than that of DCF approaches the maximum throughput as the number of

active stations increases.

It also guarantees fairness among active stations.

Previous work

Kwon et al. proposed a fast collision recovery (FCR) algorithm.

FCR uses a smaller sized contention window (CW) and reduces the value of backoff counter (BC) exponentially when idle slots are detected consecutively for a fixed number of times.

Whenever a station detects the transmission of another station, it increases the value of its CW and picks the value of BC randomly.

Protocol Specification – Terminology (1)

Service Period (SP)A period during which each active station

transmits frames in round-robin manner.

Join Period (JP)New stations can join during this period.The duration of JP (AD) is fixed and known to

all stations.

Protocol Specification – Terminology (2)

Backoff Counter (BC) It linearly depends on the number of active stations. It is decreased by one

for each DIFS (50 microsec) after a successful transmission or for each ACKTimeout after sending a data frame or for each idle slot time (20 microsec) after DIFS or ACKTimeou

t. The station transmits data when the BC becomes zero

DATA DIFSACKSIFS

DATA ACK timeout

Slot1 Slot2

5 4 3 25 4 3

Slot1 Slot2

5 4 3 24 35

5 5 5 5 5 5

5 5

BC:

BC:

Protocol Specification – Terminology (3)

The Number of Active Stations (NS) It increments NS by one when it detect a transmission Each station sets the NS to zero when the BC is reset.

Protocol Specification – State Transition Diagram

Protocol Specification – STANDBY State (1)

When a station enters the STANDBY state, it estimates the number of active stations by just counting the number of transmissions between two consecutive JPs.

How can a station to identify a JP ? If it detects idle time that is equal to AD, it

assumes that this instant is the end of the JP.

New node’sview

AD=5

SP JP SP

Protocol Specification – STANDBY State (2)

The station sets its NS as the number of active stations and updates NS continuously until it enters the JOIN state.

If a station in this state has data to send and knows the number of active stations, it enters the JOIN state at the end of the JP.

New node’sview

SP JP SP JP

The end of the JP

AD = 5

The time duration of the column where all square are white = SlotTime (20 microsec)The time duration of the column containing a grey sqaure= Tdata + SIFS (10 microsec) + Tack + DIFS (50 microsec) if successful= Tdata + ACKTimeout (30 microsec) otherwise

Protocol Specification – STANDBY State (3)

Protocol Specification – JOIN State (1)

Immediately after the end of the current JP, the station sets BC = NS + X where X = 0,1, … , AD-1

Purpose: avoid collisions among several new stations

NS = 0

The station updates NS continuously until BC becomes 0.

Protocol Specification – JOIN State (2)

Protocol Specification – JOIN State (3)

Once a station receives ACK in JP, it jumps to the ACTIVE state.

BC and NS are reset: BC = NS + 1 + AD – X

Purpose: A slot alignment operation to ensure that there are no idle slots in between active stations.

NS = 0

The latest station which enters the ACTIVE state becomes the last sender in the SP.

Protocol Specification – JOIN State (4)

ServicePeriod

Protocol Specification – JOIN State (5)

More than one new stations

Protocol Specification – JOIN State (6)

If the number of stations in the JOIN state is greater than AD, they will compete with each other to occupy one of AD slots during JP.

New stations which failed to transmit a data frame during the JP return to a STANDBY state.

Protocol Specification – ACTIVE State

An ACTIVE station uses two variables, BC and NS, to transmit a data frame without contention.

After transmitting a data frame, it sets BC to (NS + AD) and resets NS to zero

Protocol Specification – ACTIVE State

Protocol Specification – ACTIVE State The ACTIVE state is composed of two sub-

states: ACTIVE1 and ACTIVE2.

Protocol Specification – ACTIVE State When a station enters the ACTIVE state, it goes

into the ACTIVE1 state.

Protocol Specification – ACTIVE State When the sender station notices a collision by

observing the absence of ACK, the sender station goes into the ACTIVE2 state.

It tries to send the data frame in the next SP.

Protocol Specification – ACTIVE State If a station in the ACTIVE2 state suffers from a

collision again during the next SP, it enters the STANDBY state.

Protocol Specification – ACTIVE State When a station in the ACTIVE2 state sends a data frame successfully

during the next SP, it goes into the ACTIVE1 state again.

Protocol Specification – ACTIVE State The reason for using two sub-states:

Provide TCF with more robustness for possible hidden stations or other stations which are not under TCF.

Protocol Specification – ACTIVE State

If a station in the ACTIVE state has no more data to send, it will exit the ACTIVE state and transit to the STANDBY state.

Protocol Specification – ACTIVE State

If several station exit concurrently, a station in the JOIN state may determine the start of JP incorrectly and may send a data frame during SP.

It is possible that collisions may occur between stations in ACTIVE state and stations in the JOIN state.

3 2 12 11 10 4 3 2 1 059 8 67

NS=13BC=13+5=18

JP JPSPNew node’sview

Active node’sview

5 active actions exitAD=5

SP JP SP

SP

4 3 2 1 0

1 0 17 1615151413

NS=3, X = 2BC=3+2-1=4

Protocol Specification – ACTIVE State

Collision resolution scheme: Stations in the ACTIVE1 state transit to the

ACTIVE2 state. Stations in the ACTIVE2 state transit to the

STANDBY state. Stations in the JOIN state return to the

STANDBY state.

Protocol Specification – Deactivation

Performance Evaluation – Setup (1)

Several simulations are performed using ns-2 (version 2.26) simulator

TCF was compared with DCF and FCR.

FCR (CWmin=3) is selected because it is implemented distributively and provides high throughput due to a small CWmin and exponential reduction of BC.

Performance Evaluation – Setup (2)All the 802.11b stations are in a region of

70 meters x 70 meters.

The transmission range of each station is 100 meters.

Each station always has enough data to send to one of the other stations selected randomly

Performance Evaluation - Setup (3)The bit error rate (BER) is 10−5, which is th

e worst environment for an Orinoco PC card.

The traffic sources send data at a constant bit rate (CBR) and the size of a data packet is 1000 bytes. Tricky setting!

Time taken for simulations = 100 sec

Performance Evaluation – Overall Throughput (1)

Obtain a maximum throughput of DCF as a reference point There are two stations(a sender and a receiver). The sender station always has data to send. The propagation delay is negligble. The wireless network is lossless.

Performance Evaluation – Overall Throughput (2)

Obtain a maximum throughput of DCF Expected time consumed to send a data frame successf

ully (Ttransmit)

= DIFS + TE[BC]+Tdata +SIFS + Tack

where TE[BC] is the expected contention period, Tdata and Tack represent the transmission times of a data frame and an ACK respectively.

Throughputmax= TotalUsefulDataSize / Ttransmit

where TE[BC] is set to zero

i.e Ttransmit= DIFS +Tdata +SIFS + Tack

Throughputmax = 6.4617 Mbps

Performance Evaluation – Overall Throughput (3)

Performance Evaluation - Delay

where n is the number of active stations, and Xi is the measured throughput of flow i.

Performance Evaluation - Fairness

Performance Evaluation – Load Variation

My Personal Comments

Simulation scenarios are not comprehensive. The value of the constant bit rate (CBR) is not given in

this paper. The value may be “tailor-made” to produce expected

results. If the outgoing data arrival rate is much smaller than the

data transmission rate, a sender needs to join and exit frequently. The overhead incurred may increase the delay and hence decrease the overall throughput.

Performance under variable bit rate (VBR) traffic is not evaluated.

No analytical model.

Future work

Extend the operation of TCF in the environment where there are hidden stations.

Study delay-stringent versions of TCF to cope with VBR multimedia traffic.

Conclusions TCF is a simple and distributed MAC scheme for

IEEE 802.11 network to increase throughput using the information on the number of active stations.

It eliminates the contention by increasing BC to exceed the number of active stations and giving each station the opportunity to transmit a data frame in a round robin fashion.

It can be used in places where there is no hidden station such as conference rooms and coffee shops.