1. Evaluation of RR over EDCF
Month 2000
doc.: IEEE /xxx
May 2002
Evaluation of RR over EDCF
Author:
Matthew Sherman
AT&T Labs - Research
180 Park Avenue
Florham Park, NJ
Date: May 13, 2002
Matthew Sherman, AT&T Labs - Research
John Doe, His Company

2. Month 2000
doc.: IEEE /xxx
May 2002
Purpose of Document
Report on feasibility study conducted by AT&T concerning replacing CCI with RR over EDCF
Matthew Sherman, AT&T Labs - Research
John Doe, His Company

3. Background Numerous suggestions to replace CC/RR with RR over EDCF
Month 2000
doc.: IEEE /xxx
May 2002
Background
Numerous suggestions to replace CC/RR with RR over EDCF
Large simulation base for EDCF
For example 01/019, 01/133, 01/409, 01/525, 01/613
Large simulation base for CC/RR
For example 01/571r0, 02/223r1, 02/303, 02/304
No set of simulations that combine both EDCF and CC/RR
No evidence that RR over EDCF would or would not effective
Matthew Sherman, AT&T Labs - Research
John Doe, His Company

4. MAC Model Started with code from 02/304 Added EDCF parameters
May 2002
MAC Model
Started with code from 02/304
AT&T date stamp 4/7/02
Added EDCF parameters
Included Persistence Factor as nonstandard option
Defaults to standard DCF settings for b
Only supports one Traffic Category (TC) per station
All applications on station must use same TC
Not an issue for simulations conducted
Added RR over EDCF option to PCF Parameters
If set in AP, no CC will be sent
If set in STA, may send either during CCI or CP
Coded so that AP will always accept RR during CP
Decided to respond with Ack rather than CC during CP
Matthew Sherman, AT&T Labs - Research

5. New MAC Parameters Added DCF RR Option Added EDCF Parameters
Month 2000
doc.: IEEE /xxx
May 2002
New MAC Parameters
Added DCF RR Option
Added EDCF Parameters
Matthew Sherman, AT&T Labs - Research
John Doe, His Company

6. Simulation Model / Scenarios
May 2002
Simulation Model / Scenarios
Started with scenarios in 02/304
Uncontrolled DCF Sources (UDS)
modeled as Video traffic with standard DCF parameters
Added 2 new scenarios
Baseline RR over EDCF (No UDS)
All PCF stations use RR over EDCF
AP configured for no CCI
RR over EDCF with one UDS
Same UDS parameters as in 02/304
Used MAC code with AT&T date stamp 5/8/02
Used EDCF (Enhanced Contention) parameters shown on slide 5 for PCF stations
RR only EDCF traffic
When present UDS uses DCF default parameters
Matthew Sherman, AT&T Labs - Research

7. May 2002
Plots Collected
Full set of plots given in 02/223r1 provided here for comparison
Include CC/RR performance plots for benchmarking
Many plots not very meaningful for UDS
Particularly true for WLAN global statistics
May even be misleading since interactions between layers make MAC performance difficult to interpret
Key results will prove to be application delays
“Video” at end of scenario name in legend indicates UDS link is active
Matthew Sherman, AT&T Labs - Research

8. May 2002
Data Plots
Matthew Sherman, AT&T Labs - Research

9. Throughput (moving average of 240)
May 2002
Throughput (moving average of 240)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

10. Load (moving average of 240)
May 2002
Load (moving average of 240)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

11. Data Dropped (Note difference in Scale)
May 2002
Data Dropped (Note difference in Scale)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

12. Delay (global) (Note difference in Scale)
May 2002
Delay (global) (Note difference in Scale)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

13. Control Traffic Received at AP
May 2002
Control Traffic Received at AP
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

14. Control Traffic sent by AP
May 2002
Control Traffic sent by AP
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

15. Media Access Delay at Voice STA 1
May 2002
Media Access Delay at Voice STA 1
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

16. Media Access Delay at Voice STA #19
May 2002
Media Access Delay at Voice STA #19
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

17. Voice Throughput per STA
May 2002
Voice Throughput per STA
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

18. FTP Traffic served (moving average of 240)
May 2002
FTP Traffic served (moving average of 240)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

19. HTTP Traffic served (moving average of 240)
May 2002
HTTP Traffic served (moving average of 240)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

20. FTP download time (moving average of 10)
May 2002
FTP download time (moving average of 10)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

21. HTTP Page Response Time (moving average of 10)
May 2002
HTTP Page Response Time (moving average of 10)
No UDS link
One UDS link
Matthew Sherman, AT&T Labs - Research

22. Data Analysis - Plots Key differentiators are upload / download times
May 2002
Data Analysis - Plots
Key differentiators are upload / download times
As with alternatives suggested in 02/223r1, CC/RR maintains substantial voice delay advantage even without UDS
Since FTP traffic dominates, FTP upload / download times most important
RR over EDCF comparable to CC/RR without UDS
RR over EDCF highly degraded with UDS
UDS has no obvious effect on CC/RR
HTTP traffic suffers for UDS with RR over EDCF as well
Given existing plots, summary statistics seem unnecessary at this time
Matthew Sherman, AT&T Labs - Research

23. May 2002
Conclusions
Uncontrolled DCF Sources (UDS) significantly degrade performance of RR over EDCF
CC/RR is not significantly degraded by UDS
CC/RR voice performance is substantially better than for RR over EDCF
Typically 2:1
Uncontrolled DCF sources are an important issue for Managed LANs
AT&T implementation of protocol shows that CC/RR complexity is not an issue
Don’t need RR over EDCF to save on complexity
Managed WLANS are critical target market area
Cannot afford to adopt protocol such as RR over EDCF which strongly degrades with UDS
Matthew Sherman, AT&T Labs - Research