1. PCF vs. DCF: Limitations and Trends
May 2000
doc.: IEEE /yyy
January 2001
PCF vs. DCF: Limitations and Trends
Sunghyun Choi
Philips Research-USA
Briarcliff Manor, New York
Sunghyun Choi, Philips Research
Sunghyun Choi, Philips Research

2. Outline Goal Validation of simulation models of .11a/b Results of .11b
January 2001
Outline
Goal
Validation of simulation models of .11a/b
Results of .11b
Results of .11a
Conclusion
Sunghyun Choi, Philips Research

3. Goal Maximum throughput via simulation
January 2001
Goal
Maximum throughput via simulation
Throughput = Traffic (i.e., MSDU from MAC to high at the receivers) / Time
To find and compare the limitations and general trends of PCF (or polling) and DCF (or DCF-variants)
To get some idea about the limitations due to the underlying PHYs
Sunghyun Choi, Philips Research

4. Simulation Validation
January 2001
Simulation Validation
From IEEE /055: “802.11a and b max throughput for simulation model conformance”
Two STAs: one sender and one receiver
Assumptions
Infinite traffic from sender
No channel error
By varying the MPDU size
Sunghyun Choi, Philips Research

5. Analysis vs. Simulation
January 2001
Analysis vs. Simulation
Sunghyun Choi, Philips Research

6. Assumptions No EDCF nor EPCF One BSS; an AP; number of STAs
January 2001
Assumptions
No EDCF nor EPCF
One BSS; an AP; number of STAs
Infinite amount of traffic from each source
Beacon interval: 20 msec
18 msec CFP; 2 msec CP
MPDU size: 1500 bytes
No channel error
Both b and a
Sunghyun Choi, Philips Research

7. Considered Cases/Scenarios
January 2001
Considered Cases/Scenarios
DCF
# of sources = # non-AP STAs
PCF (uplink only)
No traffic from AP
# of sources = # of non-AP STAs
Downlink only case will achieve the same
PCF (both uplink/downlink)
Infinite traffic from AP to each STA
# of sources = 2 * # of non-AP STAs
Sunghyun Choi, Philips Research

8. Considered Cases/Scenarios
January 2001
Considered Cases/Scenarios
After header overhead:
Transmission rate excluding the PLCP header/preamble and MAC header overheads
MPDU size / Transmission time
PHY rate
Transmission rate at the PHY
e.g., 11 Mbps for .11b and 54 Mbps for .11a
Sunghyun Choi, Philips Research

9. 802.11b Data at 11Mbps and ACK at 2 Mbps January 2001
Sunghyun Choi, Philips Research

10. Discussion Non-small header overheads Throughput of DCF
January 2001
Discussion
Non-small header overheads
Throughput of DCF
Depends on # of sources (or STAs)
Throughput of PCF
Does NOT depend on # of sources
Does depend on the traffic pattern
Gains a lot via piggybacked CF-ACKs
Pretty high for “up/downlink” case considering the CFP size (18 msec out of 20 msec)
Sunghyun Choi, Philips Research

11. 802.11a Data at 54 Mbps and ACK at 24 Mbps January 2001
Sunghyun Choi, Philips Research

12. Discussion Relatively smaller header overheads Throughput of DCF
January 2001
Discussion
Relatively smaller header overheads
Throughput of DCF
Decreases faster as # of sources increases
Due to smaller SWmin value?
Throughput of PCF
Smaller relative to the “after header overhead” compared to .11b
Why? Relatively large SIFS (16 usec) time; losing about 3 Mbps due to SIFS time
Sunghyun Choi, Philips Research

13. January 2001
Conclusion
PCF achieves higher maximum throughput by up to 30% over DCF
Piggybacked CF-ACK is always good with PCF
Throughput of .11a is smaller than expected
Sunghyun Choi, Philips Research

14. January 2001
Conclusion (cont.)
Note that throughput will vary depending on the frame sizes
This will make the QoS management including admission control less easy
However, PCF (or polling) still gives you more predictable performance
EPCF can be better than PCF!
Sunghyun Choi, Philips Research