Transcript Experimental Measurement of VoIP Capacity in IEEE 802.11 WLANs

Experimental Measurement
of VoIP Capacity in IEEE
802.11 WLANs
Sangho Shin
Henning Schulzrinne
Motivation and Goal


Check the VoIP capacity using wireless
cards and compare it with theoretical
and simulation results
Identify all factors that affect the VoIP
capacity in experiments and simulations
Outline





Theoretical capacity for VoIP traffic
VoIP capacity via simulations
VoIP capacity via experiments
‘Hidden factors’ that affect experiments
and simulations
Conclusion
Theoretical Capacity
Packetization interval
MAC
1
2
3
backoffPLCP
…….
N
1
2
MAC IP UDP Voice
DIFS Tb
…….
3
N
PLCP ACK
SIFS
Tt
Packetization Interval (ms)
N max
P

= 15 calls
Tt  2  Tb
Capacity (calls)
PLCP = Physical Layer Convergence Procedure
parameters
value
Voice codec
64 kb/s
Packet size
160B
Packetization interval
20ms
Transport layer
UDP
PHY data rate
11 Mb/s
RTS/CTS
No
Simulation setup
QualNet simulator v3.9
parameters
IEEE 802.11b
value
Voice codec
G7.11
(64 kb/s)
Packet size
160B
Packetization interval
20ms
Transport layer
UDP
PHY data rate
11Mb/s
RTS/CTS
No
WIFI
WIFI
WIFI
WIFI
WIFI
Ethernet-Wireless
Simulation results
Delay  90th percentile
Retry rate  average
Number of VoIP sources
Capacity
Experiments
NJ Rutgers University
Experiments
80 ft
70 ft
Atheros
Intel
Experimental setup
IEEE 802.11b
param
client
client
client
client
client
AP
clients
client
client
client
client
client
client
client
client
client
value
Voice codec
G7.11
(64 kb/s)
Packet size
160B
Packetization interval
20ms
Transport layer
UDP
PHY data rate
11Mb/s
RTS/CTS
No
Experimental results
Capacity
Comparisons
Simulation
Delay
Downlink delay is larger than
uplink delay
Very low
 Increases
sharply
Retry
rate
Experiments
Gradually
increases
Uplink retry rate is higher than
downlink retry rate
Uplink:2~5%
Downlink:1~3
%
Uplink:7~11%
Downlink:4~7
%
Factors








ARF (Auto Rate Fallback)
Preamble size
PHY data rate of ACK frames
Offset of VoIP traffic start time
Signal strength
Scanning APs
Retry limit
Network buffer size
ARF

ARF (Auto Rate Fallback)

PHY data rate are automatically changes




When frame loss is caused by bad link quality, it helps
When frame loss is caused by congestion, it makes worse


Transmission failure decrease rate
Successful transmission  restore the rate
No way to tell the reason for frame losses
Problems



The effect varies according to algorithms
Turned off in simulations
Turned on in wireless cards

The algorithms are mostly implemented in drivers
ARF
ARF=AMRR
(Adaptive
Multi-Rate
Retry)
Preamble size
PLCP MAC IP UDP Voice




IEEE 802.11b : long and short preamble
Long
Short
Preamble size
144 us
72 us
Header size (b)
48 bits
48 bits
Header coding rate
1 Mb/s
2 Mb/s
Header size (us)
48 us
24 us
Total size (us)
192 us
96 us
Portion in a VoIP (size)
9%
6%
Portion in a VoIP (time)
53%
36%
Long preamble
PLCP Preamble
144us
48bits = 48us
192us
Short preamble
PLCP Preamble
72us
PLCP Header
48bits = 24us
96us
QualNet, NS-2  Long preamble
Atheros + MadWifi driver  Short preamble
Theoretical capacity with the long preamble = 12 calls
PLCP = Physical Layer Convergence Procedure
PLCP Header
Preamble size
PHY data rate for ACK frames

ACK frames

Required for ARQ
PLCP
MAC
(Automatic Repeat-reQuest)
Type : 01 Subtype 1101
14B
2Mb/s 152 us = 57% of a VoIP packet
11Mb/s106 us = 39% of a VoIP packet

Theoretical VoIP Capacity using 11 Mb/s for ACK
frames  16 calls
PHY data rate of ACK frames
Offset of VoIP traffic startime
Packetization interval
Application layer
1
2
3
4
1
2
3
4
Offset
SIFS DIFS
MAC layer
data
VoIP source 1
1
1
VoIP source 2
2
2
VoIP source 3
3
3
VoIP source 4
4
4
MAC layer
ACK
1
2
collisions
backoff
3
data
4
1
2
3
4
Offset of VoIP traffic start time
Uplink retry rate
≈ 20 ms / (15 * 2)
650us > 31 x 20us
Offset of traffic start time (us)
Simulation results with 15 VoIP sources
Key Factors








Fixed
ARF (Auto Rate Fallback)
Short
Preamble size
PHY data rate of ACK frames 2Mb/s
Offset of VoIP traffic start time Randomized
Signal strength
Scanning APs
Retry limit
Network buffer size
Signal strength
Scanning APs

Scanning APs
Probe request (broadcast)
client





Probe response (unicaset)
AP
When signal strength decreases below a threshold
When the retransmission rate increases above a threshold
Regularly (e.g. once per 30 seconds)
Hard to determine the algorithms
Problems


Management frames have a higher priority than data frames
 causes delay
Increases the traffic  make channels congested

1 probe request and 1 ~ 2 probe responses per channel
Scanning APs
Retry limit

Wireless nodes retransmit frames until the number of
retransmission reaches the retry limit

Long retry limit (4)

For the packets whose size is bigger than the RTS threshold
Short retry limit (7)
 For the packets whose size is smaller than or equal to the RTS
threshold


Effect

More retransmissions


Might reduces packet loss, but increases congestion
Less retransmissions

Increases the packet loss
Retry limit
Network buffer size

Packet loss happens mostly because of the
buffer overflow at the AP




Small buffer  increase the packet loss
Bigger buffer  reduces packet loss, but increase
the delay
Buffer size needs to be big enough to avoid the
effect
Simple static queuing analysis
Dmax
(B)
B Buffer size
   avg Average transmission time of aupacket
= 2ms
S
D = 60ms
Packet size
Maximum queuing delay (ms)
S = 200B
B = 5.8KB < 10KB MadWifi
Conclusion

Need to consider the following factors when
measuring the VoIP capacity experimentally








ARF
Preamble size
PHY data rate of ACK frames
Offset of VoIP traffic start time
Scanning APs
Retry limit
Network buffer size
By adjusting all the factors, we can achieve
the same experimental, simulation,
theoretical capacity
Thank you!