1. CC/RR Performance Evaluation - Revisited
Month 1998
doc.: IEEE /xxx
March 2002
CC/RR Performance Evaluation - Revisited
Javier del Prado and Sunghyun Choi
Philips Research USA
Briarcliff Manor, New York
J. del Prado and S. Choi, Philips

2. March 2002
Simulation Goal
In 01/571r0 the results showed a better performance with CC/RR compared to straight PCF
Straight PCF does poll all the stations
These results were not enough to validate the CC/RR mechanism
HCF should not implement a straight polling, so it is not a good comparison
We show new scenarios with more simulation results for further discussion
J. del Prado and S. Choi, Philips

3. Simulation Model Used the simulation model provided by AT&T
March 2002
Simulation Model
Used the simulation model provided by AT&T
The results in 01/571 were obtained with the same simulation model
Modified to support new scenarios
It supports:
DCF
PCF (or HCF without TXOP, working in CFP only)
CC/RR
Direct STA-to-STA transmission
J. del Prado and S. Choi, Philips

4. March 2002
Simulation Scenarios
1 AP, 6 voice STAs and 23 FTP heavy and HTTP heavy STAs
The application parameters are taken from 01/571r1.
All the application start after 100 sec from the simulation start.
The WLAN parameters are the same as in 01/571r1.
Beacon interval of 20 ms with CFP of 18 ms
J. del Prado and S. Choi, Philips

5. March 2002
Simulation Scenarios
J. del Prado and S. Choi, Philips

6. March 2002
Simulation Scenarios
J. del Prado and S. Choi, Philips

7. Simulation Scenarios 4 different simulation scenarios:
March 2002
Simulation Scenarios
4 different simulation scenarios:
Standing Poll (legacy PCF)
All STA & AP set for Standing Poll
All STA are polled using “Round Robin”
One STA is polled continuously while it has more data
The CFP ends when all STA have been polled
CC/RR
All STA & AP set for CC/RR
CC Parameters chosen from 01/571r0
Polling list updated using RR information
The STAs are polled while enough time in the CFP
Simulated in 01/571r0
J. del Prado and S. Choi, Philips

8. Simulation Scenarios New simulation scenarios:
March 2002
Simulation Scenarios
New simulation scenarios:
3) Polling only voice stations and downlink traffic to voice stations during CFP
Data downlink traffic transmitted during CP
After polling all voice stations, CFP ends
“Round Robin” polling
A voice STA is polled while it has more data to send
4) Polling voice stations and all downlink traffic during CFP
Uplink data traffic is transmitted during CP
J. del Prado and S. Choi, Philips

9. Throughput (moving average of 240)
March 2002
Throughput (moving average of 240)
J. del Prado and S. Choi, Philips

10. Load (moving average of 240)
March 2002
Load (moving average of 240)
J. del Prado and S. Choi, Philips

11. March 2002
Data Dropped AP
J. del Prado and S. Choi, Philips

12. March 2002
Delay (global)
J. del Prado and S. Choi, Philips

13. Delay (II) (without Standing Poll)
March 2002
Delay (II) (without Standing Poll)
J. del Prado and S. Choi, Philips

14. Control Traffic Received at AP
March 2002
Control Traffic Received at AP
J. del Prado and S. Choi, Philips

15. Control Traffic sent by AP
March 2002
Control Traffic sent by AP
J. del Prado and S. Choi, Philips

16. Media Access Delay at Voice STA 1
March 2002
Media Access Delay at Voice STA 1
J. del Prado and S. Choi, Philips

17. Media Access Delay at Voice STA #19
March 2002
Media Access Delay at Voice STA #19
J. del Prado and S. Choi, Philips

18. Voice Throughput per STA
March 2002
Voice Throughput per STA
J. del Prado and S. Choi, Philips

19. FTP Traffic served (moving average of 240)
March 2002
FTP Traffic served (moving average of 240)
J. del Prado and S. Choi, Philips

20. HTTP Traffic served (moving average of 240)
March 2002
HTTP Traffic served (moving average of 240)
J. del Prado and S. Choi, Philips

21. FTP download time (moving average of 10)
March 2002
FTP download time (moving average of 10)
J. del Prado and S. Choi, Philips

22. HTTP Page Response Time (moving average of 10)
March 2002
HTTP Page Response Time (moving average of 10)
J. del Prado and S. Choi, Philips

23. HTTP Object Response Time (moving average of 10)
March 2002
HTTP Object Response Time (moving average of 10)
J. del Prado and S. Choi, Philips

24. March 2002
Conclusions (I)
The scenario was designed to demonstrate the utility of CC/RR. However, comparable performance was achieved without CC/RR.
CC/RR mechanism may even degrade the overall performance in some cases due to the unnecessary overhead.
Parameterized QoS streaming will not need the CC/RR mechanism.
Resource request can be done via QoS Null or another management frame.
J. del Prado and S. Choi, Philips

25. March 2002
Conclusions (II)
We believe that the performance w/o CC/RR can be better via more intelligent scheduling
Note that …
HC can poll voice STAs during the CP (not implemented in the current simulation model)
HC can get information in the QoS control field of data frames (not implemented in the current simulation model)
CC/RR complexity is not justified!
J. del Prado and S. Choi, Philips