You are on page 1of 28

January 2001

doc.: IEEE 802.11-02/065r0

802.11g MAC Analysis


Menzo Wentink
mwentink@intersil.com

Ron Brockmann
rbrockma@intersil.com

Maarten Hoeben
mhoeben@intersil.com
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

802.11g MAC Related Settings


The following parameters are used:
802.11b aSIFSTime aSlotTime aCWmin 10 usec 20 usec 31 slots 802.11a 16 usec 9 usec 15 slots 802.11g 10 usec 20 usec 15 slots

A 6 usec silence period is added to OFDM frames, to mitigate for the 16 usec OFDM SIFS ACK frames shall be sent at a Basic Rate or PHY mandatory rate The RTS Threshold can be dynamically set by a link optimization algorithm, or by an information element in the beacon
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Recommendation: SIFS 10 usec


OFDM requires a 16, not 10 usec RX-TX turnaround This is solved in CCK-OFDM by adding a 6 usec postamble to the packet, effectively extending the SIFS for the receiver The transmitter is active longer than necessary, and the TX-RX turnaround time available is significantly reduced Recommendation: add a 6 usec silence period is added to each OFDM frame, with the same function as the CCK-OFDM postamble
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Recommendation: Slot Time 20 us


When 802.11 DS was defined, a 20 us slot was equivalent to 5 bytes at the highest rate of 2 Mbit/s Today, 20 us can transfer 135 bytes at 54 Mbit/s ! Backoff slots are very expensive this favors bursting techniques in PCF and TGe HCF Slot time is part of the definition of PIFS and DIFS affecting core MAC/TGe behaviours, and cannot be changed without significant coexistence issues
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Recommendation: CWmin 15
High cost of slot time calls for shorter backoff window 802.11a uses CWmin 15 Extensive simulations show CWmin 15 gives markedly higher overall performance in all typical scenarios than CWmin 31 802.11g nodes operating in full 802.11b backward compatibility mode (not using the 802.11g rates) should comply with 802.11b and use CWmin 31 For .11g+e products, CWmin can be overruled
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

ACK Rates
It is desired to transmit OFDM ACK frames in response to OFDM DATA frames because they are substantially more efficient Section 9.6 of 802.11-1999 and 802.11b contradict on whether this is required/forbidden when the Basic Rates do not include OFDM rates in a mixed environment Recommendation: clarify section 9.6 to support the use of OFDM Mandatory rates in response to OFDM frames even if they are not part of the Basic Rate Set as described in 02/xxx
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

RTS Threshold
RTS/CTS is used to protect OFDM frames in a mixed b/g environment Can either be enabled/disabled statically by MIB variable, or a dynamic link optimization algorithm can be used Perhaps, a Recommended Practice can be defined Legacy 802.11b STAs do not have to use RTS/CTS, unless required to optimize the link for hidden nodes or excessive collision scenarios
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Analysis of MAC Performance


DCF Performance Mixed b/g without RTS/CTS Mixed b/g with RTS/CTS, Cwmin 31 Mixed b/g with RTS/CTS, Cwmin 15 Migration rom Legacy to Pure OFDM Pure OFDM, TCP DCF Efficiency, CWmin 15/31 Pure OFDM, UDP DCF Efficiency, CWmin 15/31 TGe QoS Bursting TGe QoS Video Scenario
Brockmann, Hoeben, Wentink (Intersil)

Submission

January 2001

doc.: IEEE 802.11-02/065r0

DCF Performance

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Average Frame Tx Durations


Durations for a 1500 Byte TCP frame transmission

ofdm 24 cck-ofdm 24 pbcc 22 rts-cts ofdm 24 cck (11b)


0 150 300 450 600 750 900 1050 1200 1350 1500 1650 1800

rts cts data ack av. backoff 15 av. backoff 31

Transmission Time (usec)

*) RTS CTS OFDM features cheap collisions (cost of one RTS) and built-in hidden node protection
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Throughput Comparison for 24/22 Mbps


802.11g Performance (22/24 Mbps, CWmin = 15)
16 15 14 13 12 11 10

Throuhput 8 (Mbps)
7 6 5 4 3 2 1 0

cck11

pbcc22

ofdm rts/cts

cck-ofdm

ofdm

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Mixed b/g

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Mixed b/g without RTS/CTS


Performance in a mixed scenario, without RTS/CTS (802.11g / legacy)
10 9 8

Throughut (Mbps)

the aggregate throughput goes down


7 6 5 4 3 2 1 0 0 1 2 3 4 5 6 7 8 9 10 11

The unprotected OFDM packets collide with legacy CCK. The OFDM TCP flows are starved.
Node 1 (802.11g) Node 2 (802.11g) Node 3 (legacy) Node 4 (legacy) Aggregate

4 legacy nodes

2 OFDM nodes without RTS/CTS + 2 legacy nodes

the throughput of the legacy nodes goes up

Time (sec)

The throughput of OFDM nodes diminishes, because 12 OFDM yields for CCK, but not v.v.

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Mixed b/g with RTS/CTS, CWmin 31


Perormance in a mixed scenario with RTS/CTS (CWmin = 31 for 802.11g)
10 9 8

the aggregate throughput goes up Protected OFDM transmissions nicely mix with legacy 2 OFDM nodes with RTS/CTS 2 legacy nodes
Node 1 (802.11g) Node 2 (802.11g) Node 3 (Legacy) Node 4 (Legacy) Aggregate

Throughput (Mbps)

7 6 5

4 legacy nodes
4 3 2 1 0 0 1 2 3 4 5

The throughput of OFDM and legacy goes up by same amount due to fairness of DCF. RTS/CTS-protected
6 7 8 9 10 11 12

Time (sec)
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

DCF Fairness
For equal CWmin, throughput increase is distributed over all nodes!
DCF gives each node equal number of transmit opportunities, regardless of their data rate Legacy 802.11b frame transmissions are longer and they hog media time with their inefficient modulations Aggregate throughput increases but less than expected

By using a smaller CWmin, TGg nodes can get higher priority


Since their transmissions are shorter, total time spent on the media is comparable to legacy nodes
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Mixed b/g with RTS/CTS, CWmin 15


Performance in a mixed scenario, with RTS/CTS (CWmin = 15 for 802.11g)
10 9 8

the aggregate throughput goes up RTS/CTS-protected OFDM transmissions nicely mix with legacy 2 OFDM nodes with RTS/CTS + 2 legacy nodes
Node 1 (802.11g) Node 2 (802.11g) Node 3 (legacy) Node 4 (legacy) Aggregate

Throughput (Mbps)

7 6 5 4 3 2 1 0 0 1 2 3 4 5 6 7 8 9 10 11

4 legacy nodes

the throughput of OFDM nodes goes up, because of more efficient transmissions and smaller CWmin.
12 the legacy throughput

Time (sec)

levels

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Migration from Legacy to 802.11g

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Migration to 802.11g from legacy


TCP performance during migration to 802.11g (CWmin = 15, OFDM 36 for 802.11g nodes)
20 18 16

aggregate throughput
4 g-nodes w/o rts/cts 4 g-nodes 3 g-nodes 1 b-node

Throughput (Mbps)

14 12 10 8 6

Node 1 Node 2 Node 3 Node 4 Aggregate

4b
4 2 0 0 5

2 g-nodes 2 b-nodes

Individual throughputs
10 15 20 25

OFDM and legacy CCK transmissions are mixed.


Submission

Time (sec)

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Pure OFDM UDP Performance Comparison

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Performance in relation with CWmin (1)


Performance compared for CWmin = 15 and CWmin = 31
26 25 24

CWmin = 15 CWmin = 31
CWmin = 15 CWmin = 31

Throughput (Mbps)

23 22 21 20 19 18 17 16 1 2 3 4

# backlogged contenders
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Performance in relation with CWmin (3)


CWmin 15 vs. 31
30

CWmin = 15
25

Throughput (Mbps)

20

CWmin = 31

15

CWmin = 15 CWmin = 31

10

0 0 5 10 15

# backlogged contenders
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Pure OFDM TCP Performance Comparison

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Throughput comparison for TCP


Contention Window comparison

25

20

15

Throughput (Mbps)
10

CWmin = 31 CWmin = 15

Rate = 36

Rate = 54

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

802.11e QoS Scenarios

Submission

Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Migration with 802.11e HCF Bursting


802.11e/g migration scenario
20 18 16 14

Aggregate throughput
3 g-nodes (CFBs) 1 b-node

4 g-nodes

Mbps

12 10 8 6 4 2 0 0 5
4 b-nodes

2 g-nodes (CFBs) 2 b-nodes

Node 1 Node 2 Node 3 Node 4 Aggregate

Individual throughputs
10 sec 15 20 Legacy throughput levels
Brockmann, Hoeben, Wentink (Intersil)

Throughput for g-nodes rises sharply


Submission

January 2001

doc.: IEEE 802.11-02/065r0

Streaming video with 802.11e/g


2x 12 Mbps Video over 802.11g, in legacy environment, with 802.11e HCF CAPs
30

25

Throughput (Mbps)

aggregate throughput
20 Aggregate 15 Node 1 (.11g video) Node 2 (.11g video) Node 3 (.11b legacy) 10 Node 4 (.11b legacy)

2x 12 Mbps video

0 0 5 10 15 20

Time (sec)
Submission

no starvation of background
Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Simulation Environment
Network Simulator (NS)
from University of California 802.11 added by Carnegie Mellon 802.11e EDCF added by Atheros

We added
802.11g PHY (next to 11b PHY) Dynamic Rate selection and duration calculation 802.11e Contention Free Bursting

Typical simulation setup


4 stations (b or g) and 1 AP (g)
Submission Brockmann, Hoeben, Wentink (Intersil)

January 2001

doc.: IEEE 802.11-02/065r0

Conclusions
Mixed 802.11b/g operation increases network throughput Pure 802.11g operation is efficient TGe enhancements work for mixed and pure g networks; provide greater MAC efficiency Recommendations to be adopted
Submission Brockmann, Hoeben, Wentink (Intersil)

You might also like