November 2000 doc IEEE 802 11 00373 802
November 2000 doc. : IEEE 802. 11 -00/373 802. 11 PCF Model Progress Update Nov 2000 Author: Matthew Sherman AT&T Labs - Research 180 Park Avenue Florham Park, NJ 07932 973 -236 -6791 mjsherman@att. com Date: November 8, 2000 802. 11 PCF Model Progress Update Nov 2000 1 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Introduction / Background • Prior descriptions in: – IEEE 802. 11 -00/264 r 1 – IEEE 802. 11 -00/318 r 1 • PCF model developed by AT&T Labs – Validation still required • Jumping off point for MAC enhancements • Currently available under NDA from AT&T – May become part of OPNET Standard Library – Eventual contributed model with enhancements • No NDA for contributed model 802. 11 PCF Model Progress Update Nov 2000 2 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Mods Since Sept ‘ 00 Meeting • Implemented more_data bit in frame format – PCF use only • Streamlined frame formats – New packet formats called: • Wlan_mac_pcf • Wlan_control_pcf – Added all possible fields that might be needed for data / management frames – Reduced Control Frame fields since some irrelevant • Fix to beacon header size – Off by 6 bytes • Receiver busy / SIFS fix – Receiver busy now ignored during SIFS 802. 11 PCF Model Progress Update Nov 2000 3 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Mods Since Sept ‘ 00 Meeting (Cont. ) • Fixed frame timeout for PCF side traffic – AP was not waiting long enough for traffic • EIFS fix – Was still being set in NAV for some cases • PCF Duration field fixes – Was setting to 0 rather than 32768 • AP NAV reset at start of CFP – May reconsider later • Option to reset NAV for CF_END for all BSSs or only own BSS • NS Compatible trace capability • Update to October models capability 802. 11 PCF Model Progress Update Nov 2000 4 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Mods Since Sept ‘ 00 Meeting (Cont. ) • Assorted Retry fixes – Wrong frame type, duration filed, trace info, etc. • • Added support for multi-source model “Ghost Packet” and other receiver fixes TBTT overrun fix Max packet size fix Broadcast packet CW reset fix Fragment train before Beacon Fix Various other minor fixes PCF poll flag fixes – sometimes PCF active flag was set by mistake 802. 11 PCF Model Progress Update Nov 2000 5 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Current Scenario Simulated • Based on most current “Model 3 Scenario” – Modified Video Conference packet generation • Packets were too large – Added bulk traffic stream 10 sec into simulation • Load system beyond channel capacity – Addresses assigned to favor higher priority traffic • Addresses used as AID by poll scheduler • Will update scenario to final “Model 3” once completed – MPEG sources – Finalize Video Conference and data streams 802. 11 PCF Model Progress Update Nov 2000 6 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 OPNET Scenario 802. 11 PCF Model Progress Update Nov 2000 7 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 MAC Parameters 802. 11 PCF Model Progress Update Nov 2000 8 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Traffic Sources Source Address Start Time (Sec) ON State Time (Sec) OFF State Time (Sec) Interarrival Time (Sec) Packet Size (Bytes) Stop Time (Sec) Destination Address Type of Service Stream ID 0 0. 02 3600. Uniform (0. 08333, 0. 25) 815 Never 3 0 5 0 0. 02 0. 012 0. 088 0. 001 1464 Never 4 0 6 0 10. 02 3600. Poisson (0. 005) 2300 Never 6 0 4 1 0. 02 3600. 0. 02 92 Never 2 0 0 2 0. 02 3600. 0. 02 92 Never 1 0 1 3 0. 02 3600. 0. 02 92 Never 0 0 2 3 0. 02 3600. Uniform (0. 08333, 0. 25) 40 Never 0 0 3 4 0. 02 0. 012 0. 088 0. 001 1464 Never 0 0 7 5 0. 02 3600. 0. 003012 1504 Never 6 0 8 5 0. 02 3600. Lognormal (0. 01136, 0. 00001) 1504 Never 6 0 8 6 10. 02 3600. Poisson (0. 005) 64 Never 0 0 9 802. 11 PCF Model Progress Update Nov 2000 9 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Global Stats PCF 802. 11 PCF Model Progress Update Nov 2000 DCF 10 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Global Stats • 100 values / bin in simulations – Groups of values (100) are processed at a time to generate displayed statistics • Global Stats are aggregate values across all streams • Delay is time from when packet queued in MAC until final Ack indicating successful delivery • Load is traffic presented to MAC from upper layers • Media Access Delay is time from packet queued in MAC until first access of medium • Throughput is traffic forwarded to upper layers • Packets dropped self explanatory 802. 11 PCF Model Progress Update Nov 2000 11 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Global Stats - Performance • PCF outperforming DCF from start – Some packet drops on DCF, none on PCF – PCF delays lower • See effects of Bulk data kick in at 10 sec – Large number of packet drops – Delays start getting really bad • Throughput on PCF increases slightly when Bulk data starts • Throughput on DCF seems to slightly decrease – More retrys / backoffs 802. 11 PCF Model Progress Update Nov 2000 12 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Dropped Packets Voice Audio Ack Bulk Data Audio Video Conf. MPEG Data Ack PCF 802. 11 PCF Model Progress Update Nov 2000 DCF 13 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Dropped Packets • Clearly see that DCF is dropping MPEG traffic almost from the start – Probably running out of buffer room • When Bulk data starts, DCF drops even more MPEG and bulk video – No other data affected (in first 20 sec) • Video conferencing is affected in longer simulations (>30 sec) – Seems to be function of overall rate (buffer room) • PCF drops bulk data but nothing else – Only drops lowest priority stream – Maintains all Qo. S streams 802. 11 PCF Model Progress Update Nov 2000 14 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Retransmission Attempts PCF 802. 11 PCF Model Progress Update Nov 2000 DCF 15 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Retransmission Attempts • Substantial retransmission activity for DCF – Bulk Data has no obvious effects on retry • Medium already saturated – Retransmissions consume capacity • No Retransmissions required in PCF – Obvious and Expected – Efficiency advantage over DCF 802. 11 PCF Model Progress Update Nov 2000 16 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Packet Delays Voice Audio Ack Bulk Data Audio Video Conf. MPEG Data Ack PCF 802. 11 PCF Model Progress Update Nov 2000 DCF 17 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Packet Delays • PCF packet delays very stable and well bounded on all streams but bulk data – Can maintain Qo. S – Especially true for voice – Some dependence on polling order • DCF shows substantially greater delay variation in all streams – MPEG already degraded – Video Conferencing and Audio hit hard by bulk data – Qo. S is not maintained • Streams sharing AP with Bulk data hit hardest 802. 11 PCF Model Progress Update Nov 2000 18 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 DCF Trace - Synchronized Data s -t x -t r -t r -t s -t x -t d -t d -t x -t d -t s -t x -t r -t r -t r -t 0. 078352 -Hs 0. 078464 -Hs 0. 078465 -Hs 0. 080000 -Hs 0. 080087 -Hs 0. 080088 -Hs 0. 080504 -Hs 0. 081074 -Hs 0. 081161 -Hs 0. 081162 -Hs 0. 081172 -Hs 0. 081284 -Hs 0. 081284 -Hs 6 -Hd -2 -Ni 0 -Hd -2 -Ni 5 -Hd -2 -Ni 4 -Hd -2 -Ni 2 -Hd -2 -Ni 3 -Hd -2 -Ni 1 -Hd -2 -Ni 0 -Hd -2 -Ni 1 -Hd -2 -Ni 2 -Hd -2 -Ni 3 -Hd -2 -Ni 1 -Hd -2 -Ni 0 -Hd -2 -Ni 4 -Hd -2 -Ni 5 -Hd -2 -Ni 0 -Hd -2 -Ni 2 -Hd -2 -Ni 6 -Hd -2 -Ni 2 -Hd -2 -Ni 0 -Hd -2 -Ni 1 -Hd -2 -Ni 3 -Hd -2 -Ni 6 -Hd -2 -Ni 4 -Hd -2 -Ni 5 -Hd -2 -Ni 1 -Hd -2 -Ni 3 -Hd -2 -Ni 4 -Hd -2 -Ni 2 -Hd -2 -Ni 0 -Hd -2 -Ni 5 -Hd -2 -Ni 6 -Nx 0 -Nx 5 -Nx 4 -Nx 2 -Nx 3 -Nx 1 -Nx 0 -Nx 1 -Nx 2 -Nx 3 -Nx 1 -Nx 0 -Nx 4 -Nx 5 -Nx 0 -Nx 2 -Nx 6 -Nx 2 -Nx 0 -Nx 1 -Nx 3 -Nx 6 -Nx 4 -Nx 5 -Nx 1 -Nx 3 -Nx 4 -Nx 2 -Nx 0 -Nx 5 -Nx 6 -Nx -53. 26 -Ny -53. 26 -Ny -53. 26 -Ny -53. 26 -Ny 802. 11 PCF Model Progress Update Nov 2000 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 19 0 -Md 0 -Md 0 -Md 123 -Md 123 -Md 0 -Md 0 -Md 123 -Md 123 -Md 0 -Md 0 -Md 5 -Ms 6 -Mt Wlan. C_Ack 5 -Ms -2 -Mt Wlan. C_Ack -1 -Ms 0 -Mt Wlan. C_Beac 2 -Ms 1 -Mt Wlan. C_Data 1 -Ms 2 -Mt Wlan. C_Data 0 -Ms 3 -Mt Wlan. C_Data 2 -Ms 1 -Mt Wlan. C_Data 0 -Ms 3 -Mt Wlan. C_Data 1 -Ms 2 -Mt Wlan. C_Beac -1 -Ms 0 -Mt Wlan. C_Beac 1 -Ms 2 -Mt Wlan. C_Data 1 -Ms 2 -Mt Wlan. C_Data 2 -Ms 1 -Mt Wlan. C_Ack 2 -Ms -2 -Mt Wlan. C_Ack Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 Key Trace features • NS compatible • First letter is event type • • • T: packet Tx start X: packet Tx end R: packet received D: packet drop -t indicates simulation time (rounded to microsec) -Hs is observing (reporting) station address -Na is Duration field -Md is packet destination -Ms is packet source Frame type at end of line 802. 11 PCF Model Progress Update Nov 2000 20 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 DCF Trace - Synchronized Data • Multi-packet collision is common in simulations – Prior to beacon all STA have completed post-backoff – Data frame generated in sources at same time • 3 voice calls have identical parameters • Also synchronized with Beacon • Collisions could be avoided by phasing sources to discourage simultaneous triggering • What about network synchronization – Can devices in real network become synchronized so that this becomes a common problem? 802. 11 PCF Model Progress Update Nov 2000 21 Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 DCF Trace - Delayed Beacon s -t 0. 118858 -Hs x -t 0. 119943 -Hs 0 -Hd -2 -Ni 0 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 123 -Md s -t 0. 119954 -Hs x -t 0. 120066 -Hs 4 -Hd -2 -Ni 4 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na s -t 0. 120337 -Hs x -t 0. 121451 -Hs 5 -Hd -2 -Ni 5 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 123 -Md s -t 0. 121461 -Hs x -t 0. 121573 -Hs 6 -Hd -2 -Ni 6 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na s -t 0. 121684 -Hs x -t 0. 121771 -Hs 3 -Hd -2 -Ni 3 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 123 -Md s -t 0. 121782 -Hs x -t 0. 121894 -Hs 0 -Hd -2 -Ni 0 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na s -t 0. 121965 -Hs x -t 0. 122052 -Hs 1 -Hd -2 -Ni 1 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 123 -Md s -t 0. 122062 -Hs x -t 0. 122174 -Hs 2 -Hd -2 -Ni 2 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na s -t 0. 122265 -Hs x -t 0. 123350 -Hs 4 -Hd -2 -Ni 4 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 123 -Md s -t 0. 123360 -Hs x -t 0. 123472 -Hs 0 -Hd -2 -Ni 0 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 0 -Md 4 -Ms 0 -Mt Wlan. C_Ack s -t 0. 123623 -Hs x -t 0. 124127 -Hs 0 -Hd -2 -Ni 0 -Nx -53. 26 -Ny 24. 91 -Nz 100. 00 -Ne -1. 000000 -Nl MAC -Nw --- -Na 0 -Md -1 -Ms 0 -Mt Wlan. C_Beac 802. 11 PCF Model Progress Update Nov 2000 22 0 -Md 0 -Md 4 -Ms 0 -Ms 6 -Ms 5 -Ms 0 -Ms 3 -Ms 2 -Ms 1 -Ms 0 -Mt Wlan. C_Data 4 -Mt Wlan. C_Ack 5 -Mt Wlan. C_Data 6 -Mt Wlan. C_Ack 3 -Mt Wlan. C_Data 0 -Mt Wlan. C_Ack 1 -Mt Wlan. C_Data 2 -Mt Wlan. C_Ack 4 -Mt Wlan. C_Data Matthew Sherman, AT&T Labs - Research
November 2000 doc. : IEEE 802. 11 -00/373 DCF Trace - Delayed Beacon • For DCF, Beacon is required to follow DCF rules – Must back off and defer like any other frame • In simulations backoff and deference can cause Beacon to be delayed beyond TBTT by several frames – In this case delayed 9 frames – Delayed roughly 3. 6 milliseconds • Nothing wrong, but interesting to observe – Demonstrates utility of trace 802. 11 PCF Model Progress Update Nov 2000 23 Matthew Sherman, AT&T Labs - Research
- Slides: 23