Call Admission Control in IEEE 802'11 WLANs using QPCAT

1
Call Admission Control in IEEE 802.11 WLANs using
QP-CAT

• Sangho Shin, Henning Schulzrinne
• Columbia University

2
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

3
Introduction

• Problem
• VoIP in wireless LANs has increased.
• When number of VoIP flows is beyond capacity the
  quality of all calls degrades.
• Quality degradation is linked to increase in
  downlink delay at AP.
• Solution
• Prevent number of VoIP flows going beyond
  capacity.
• Use call admission control to protect existing
  VoIP flows.
• Queue size Prediction using Computation of
  Additional Transmission (QP-CAT)

4
Introduction

• Contributions
• Verify correlation between AP queue size and
  downlink delay.
• Propose queue size of AP as metric to perform
  call admission control.
• Evaluate approach in MadWifi driver and QualNet
  simulation.

5
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

6
Related Work
7
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

8
Queue Size Delay Correlation

• Queue size at AP is used to estimate the QoS of
  all VoIP flows.
• Downlink delay represents QoS for all flows
• Correlation verified by ORBIT test-bed
  measurements and analysis.

9
Queue Size Delay Correlation
10
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

11
QP-CAT

• Best measure of queue size is after flow is
  admitted.
• Alternative is to estimate the queue size before
  flow is admitted.
• QP-CAT predicts future queue size by emulating a
  new VoIP flow before the flow is requested.

12
QP-CAT
13
QP-CAT

• A new VoIP flow is emulated using uplink and
  downlink counters.
• Counters are incremented based on packetization
  interval.
• Counters are decremented based on CAT
  computation.

14
QP-CAT

• Number of additional frames

15
QP-CAT
16
QP-CAT
17
QP-CAT
18
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

19
Test Results - Simulation

• Simulation Setup
• QualNet 2.9
• 802.11b and fixed 11Mbs rate
• RTS/CTS disabled

20
Test Results - Simulation
21
Test Results Test Bed

• QP-CAT was implemented in MadWifi device driver.
• Tested on ORBIT wireless test bed
• 20 x 20 node grid of wireless nodes
• Each node has two wireless and two ethernet
  interfaces

22
Test Results Test Bed

• MadWifi Notes
• A second wireless card was used in monitor mode
  to capture uplink and downlink frames.
• Idle time was computed based on the RX timestamp
  and the transmission time.
• Transmission time was computed using frame size /
  data rate

23
Test Results Test Bed
24
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

25
QP-CAT Extensions

• 802.11e provides priority between VoIP and TCP
  traffic.
• Modify QP-CAT to consider the unused TXOP when
  computing the idle time.

26
QP-CAT Extensions
27
QP-CAT Extensions

• Parallel execution can be used to handle multiple
  type of VoIP traffic.
• Serial execution can be used to handle multiple
  admissions that arrive close together.

28
Agenda

• Introduction
• Related Work
• Queue Size-Delay Correlation
• QP-CAT
• Test Results
• QP-CAT Extensions
• Conclusion

29
Conclusion

• QP-CAT
• The predicted downlink queue size is used as
  metric to make call admission decision.
• Queue size of new call is computed continuously.
• Simulation and experiment results show the
  prediction is close to actual queue size.
• Note
• Results for multiple traffic type not included.
• QP-CATe only considers two traffic classes but
  four are available.

30
Questions/Comments?