Short-term Fairness

for TCP Flows

in 802.11b WLANs

Authors: M. Bottigliengo. C. Casetti, C.-F. Chkdsserini. M. Mea

Presented by: Tran Thi Minh Khoa
Introduction
• Wireless Local Area Networks (WLANs) based on the IEEE 802.11
technology are becoming increasingly popular and widely deployed
• However, the growing need for Quality of Service (QoS) is
difficult to implement in distributed systems like WLANs
• The basic requirement of providing fair access to all user is
conflicting with the nature of higher-layer protocols: TCP is fair
only under certain conditions, hardly met by 802.1lb WLANs.
• Another basic requirement is the protection for short-lived TCP
flows, that are sensitive to losses during the early stages of the TCP
window growth

2
Introduction
• The main contribution of this paper is the proposal of an LLC-layer
algorithm that can be implemented at both Access Point (AP) and
Wireless Stations (WSs)
• LLC – Logical Link Control: provides an interface to higher layers,
performs flow and error control

• The algorithm aims at

• guaranteeing fair access to the medium to every user, by awarding
longer transmission opportunities to WSs that experienced short channel
failures.
• protect short-lived flows while they strive to get past the critical "small
window regime"

3
 Reference Scenario
• Consider a wireless-cum-wired
- network & traffic network
scenario
• A fixed node S is connected with an AP through a
wired link of speed R and propagation delay D
• The wireless portion is an 802.11b WLAN with N
WSs
• Each WS can directly communicate only with the
AP (focus on AP-coordinated wireless networks)
• The RTS/CTS mechanism is employed
• Assume that WSs and the AP operate at a data rate
of 11 Mbps
• At transport layer. TCP connections are established
4
between each WS and the fixed node S
Reference Scenario
- channel model

• An independent error model for each communicating pair

of node, due to
• the burstiness and time correlation of wireless errors
• WLAN may experience location-dependent channel errors

• Is represented by a three-state discrete-time Markov chain

• Errors over the channel occur in the states long bad (LB) and
short bad (SB)
• Error-free : good (G)

5
Reference Scenario
- channel model

• Average number of bad slots experienced

when the states LB and SB are entered
• and
• : transition probability for state LB(SB) to G

•• Average
  number of consecutive error-free
slots is

• Parameter k is the probability that the

Markov chain moves to state LB

6
Scheduling Algorithm Description

• Goals of scheduling algorithm

• 1st – to improve the fairness among wireless stations that
may experience location-dependent channel capacity and
errors
• 2nd – to provide short-term fairness to TCP traffic in order to
enhance the performance of short-lived flows

7
Scheduling Algorithm Description

• To achieve the goals

• At the AP LLC layer: a separate queue for each WS
associated to the AP
• WS LLC layer: only one queue is implemented
• A channel condition estimator is associated to each queue
• Transmission is allowed only for those queues whose
channel is estimated to be good

8
Scheduling Algorithm Description

• Queues attempting to access the wireless medium with a

bad channel will be allowed to transmit again when their
channel becomes good
• Upon switching bad  good:
• a queue sends to the MAC layer a maximum number of back-
to-back frames (TXburst)
• The MAC layer will then use a burst transmission fashioned
after the EDCF bursting capability of the current 802.1le draft,
to avoid contentions within the burst

9
Scheduling Algorithm Description

• The building blocks of Scheduling Algorithm are the

following:
• channel state estimation
• queue selection and service
• TXburst length setting

• The algorithm is implemented both at the AP and at WSs

• Note: if only one queue is used at each WS. the queue
selection in not needed at WSs

10
Scheduling Algorithm Description
- Channel State Estimation

• the AP estimates the channel conditions for each contending WS

• a flag is associated to each LLC queue at the AP indicating the

corresponding channel state (GOOD, BAD, PROBE)
• GOOD: whenever AP receive from the corresponding WS
• a MAC layer ACK

• a CTS fame
• an error-free data frame or RTS

• BAD: after transmission failure

• PROBE: when a configuration timeout (PTIMER) expires

11
Scheduling Algorithm Description
- Queue Selection and Service
Queue Service
Queue Selection

12
 Scheduling Algorithm Description
- Txburst Length Setting

• 1st, AP compute a fair target throughput

• 2nd, assume queue i is selected to be served,

and the Txburst length needs to be assessed.
•  The AP compares the values of Queue_Thr(i)
and Thr_Fair, and computes
• 3rd, AP includes the value of Thr_Fair in
control frames that are broadcast on the
WLAN

13
Simulation Scenario & Numerical Result
• S has 55 Mbps, propagation delay 2 ms
• no hidden stations

• Higher layer setting

• TCP connections exhibiting an On-Off behavior

• Transport layer Settings

• NewReno
• Maximum Segment Size (MSS) 1000bytes
• Congestion window: 1 – 100 segments

• Data Link layer Settings

• The RTS/CTS mechanism is employed

14
 Simulation Scenario & Numerical Result

Average per-connection throughput vs.

Average per-connection throughput vs.
number of WSs
average slot error probability

15
 Simulation Scenario & Numerical Result

Fairness index of connection with size of 27 TCP timeout prevention

segments

16
Thank you for your listening

Tran Thi Minh Khoa