Mes essag age e in M Mes essag age M e Mec echa hanism sm - - PowerPoint PPT Presentation
Potential al Pitf tfalls ls o of f the he Mes essag age e in M Mes essag age M e Mec echa hanism sm in M Mode odern 8 n 802.11 Ne Networks Wei Wang, Wa Wai Kay L y Leon eong, and Ben Leong School of Computing, National
Potential al Pitf tfalls ls o
f the he
Mes essag age e in M Mes essag age M e Mec echa hanism sm
in M Mode
n 802.11 Ne Networks
Wei Wang, Wa Wai Kay L y Leon eong, and Ben Leong
School of Computing, National University of Singapore
Wi-Fi is Ubiquitous
2
The Problem
Message in Message Mechanism (MiM)
MAC protocol ACK Interference Power Control
3
MESSAGE IN MESSAGE MECHANISM
Conventional Receiver w/o MiM
Frame A
Time RSSI
Frame B
Higher RSSI
5
Reception of Conventional Receiver
Both frames are lost Frame A
Time RSSI
Frame B
corrupted CRC check fails
Treated as noise
6
Frame B
Frame A knocked out
Message in Message (MiM)
Higher signal dominates weaker signal
Time RSSI
Frame A
Successfully Received
7
Frame A knocked out
Frame B
Interfering Frame Desired Frame
MiM is helpful
Time RSSI
8
Discarded
Successfully Received
Desired Frame Interfering Frame
MiM is helpful
Time RSSI Successfully Received
Discarded
9
Desired Frame Interfering Frame
MiM is helpful, at least no harm
Time RSSI Successfully Received
Discarded
10
Consider Aggregate MPDUs
However…
A-MPDU MAC Frames
11
However… Consider A-MPDU
Time RSSI A-MPDU Interfering frame
12
However… Consider A-MPDU
Without MiM RX: 3 Fail: 3
Time RSSI A-MPDU Interfering frame
13
Interfering frame
However… Consider A-MPDU
Without MiM With MiM RX: 3 Fail: 3 RX: 1 Fail: 5 Worse: No Block ACK
Time RSSI A-MPDU
A-MPDU gets knocked out
Key Insight: MiM can be harmful
14
Why Use A-MPDU?
the whole A-MPDU
15
SOMETIMES GOOD, SOMETIMES BAD
HOW TO EFFECTIVELY USE MIM
Our Contributions
A: Study the impact of MiM on A-MPDUs
A: Adaptive algorithm to enable/disable MiM
18
Studying the Impact of MiM
Experimental set-up
Sender Receiver Interferer
19
Studying the Impact of MiM
Experimental set-up
Sender Receiver Interferer Interfering Frame
20
Studying the Impact of MiM
Experimental set-up
Sender Receiver Interferer Interfering Frame
21
Ensure collision
Time Interferer Receiver Sender Poll
t
t is uniformly distributed Immediately Tx A-MPDU Interfering Frame
22
Duration of A-MPDU
Max duration limited by ath9k driver
Time
≈ 3.8 ms
A-MPDU
Max. 4 ms
23
A-MPDU
Size of A-MPDU (# frames)
Depends on data rate
6.5 Mbps 2 frames 26 Mbps 8 frames
Time
≈ 3.8 ms
MCS Index 1 2 3 4 5 6 7 Data Rate (Mbps) 6.5 13 19.5 26 39 52 58.5 65 Frames 2 4 6 8 12 16 18 20 Max. 4 ms
24
The Detrimental Impact of MiM
26
NUMBER OF FRAMES IN AN A-MPDU
Impact of A-MPDU size
A-MPDU of 2 frames
Time
≈ 3.8 ms Interfering Frame 60 μs
0.5
28
Impact of A-MPDU size
A-MPDU of 4 frames
Time
≈ 3.8 ms Interfering Frame 60 μs
29
Impact of A-MPDU size
A-MPDU of 4 frames
0.25
30
Impact of A-MPDU size
A-MPDU of 20 frames
Time
≈ 3.8 ms Interfering Frame 60 μs
More details in the paper
31
Frame Delivery Ratio
Time
≈ 3.8 ms Interfering Frame 60 μs
32
Interfering Frame 60 μs
Frame Delivery Ratio
0.5 0.9
Time
≈ 3.8 ms Interfering Frame 600 μs
33
Interfering Frame 600 μs
Frame Delivery Ratio
Time
≈ 3.8 ms
0.5
34
THE AIR-TIME DURATION
Air-time of Interfering Frames
Intuition: Without MiM, longer T more frames loss With MiM, T has no effect
Time
≈ 3.8 ms
T
36
How to set T
Time
≈ 3.8 ms
T
37
Increasing Frame Length
38
Increasing Frame Length
39
Increasing Frame Length
40
Increasing Data Rate
Air-time duration is what matters
41
Air-time Duration… in the Wild
42
Air-time Duration… in the Wild
Median ≈ 30 μs MAC ACK
IPv6 Neighbor Discovery Protocol
20 μs 170 μs
43
Putting it in Perspective
in the wild
Be careful what you choose Suffer a large penalty
44
USING ADJACENT CHANNELS
Channel bonding
Sender Interferer Receiver Interferer
20 MHz 40 MHz
50
Channel bonding: Case 1
Sender Receiver Interferer
51
Channel bonding: Case 2
Sender Interferer Receiver
52
Channel bonding: Case 3
Sender Interferer Receiver
53
Channel bonding: Case 4
Sender Interferer Receiver
54
Channel bonding: Case 5
Sender Interferer Receiver
55
Sender Interferer Receiver
Channel bonding
Sender Interferer Receiver
Case 1
Sender Interferer Receiver
Case 2
Sender Interferer Receiver
Case 3
Sender Interferer Receiver
Case 4
Sender Interferer Receiver
Case 5
56
Adjacent Channel Interference
Sender Interferer Receiver
58
Adjacent Channel Interference
Sender Interferer Receiver
Lesser Interference More Interference 10 dB Threshold
59
Adjacent Channel Interference
Sender Interferer Receiver
60
Adjacent Channel Interference
Sender Interferer Receiver Receiver
61
Adjacent Channel Interference
Sender Interferer Receiver
62
DECIDING WHEN TO ENABLE/DISABLE MIM
Some Definitions
Good Knock-out Bad Knock-out
Frame A knocked out Time RSSI
Desired Frame
Successfully Received
Interfering Frame
Discarded
Desired Frame Interfering Frame
Time RSSI Discarded
Discarded
64
Key Idea
Count Good KO and Bad KO Compare Good > Bad Enable MiM Yes Disable MiM* No *CATCH Cannot count with MiM disabled Periodically
65
Evaluation
Experimental Set-up Position 1 Position 2 Position 3
Desired signal is stronger Equal signal strength Interference is stronger
Sender
Campus AP
Interferer
Campus AP
66
Results w/o Adaptive MiM
Sender > Interferer MiM helpful Sender = Interferer MiM neutral Sender < Interferer MiM detrimental
67
Results with Adaptive MiM
Sender > Interferer MiM helpful Sender = Interferer MiM neutral Sender < Interferer MiM detrimental
Adaptive MiM always useful
68
In Conclusion
MiM not always helpful, can be harmful
69
Future Work
simulators like ns-3
MPDU
MPDU size
70
Thank You
QUESTIONS?
{weiwang|waikay|benleong}@comp.nus.edu.sg