Transcript WifiNetMon: Fault

Performance Analysis,
Anomaly Detection, Diagnosis & Resolution
in Wi-Fi Mesh Network
M.Tech. Thesis Presentation
June 25, 2009
by
Rahul Mittal
Under guidance of
Prof. Bhaskaran Raman
Indian Institute of Technology, Bombay
Outline
• Motivation & Problem Statement
• Prior Work and their shortcomings
• Overall Approach
– Experiment Daemon & Types of experiments
– Link Diagnosis Algorithm
•
•
•
•
IIT Bombay WiFi Mesh Network
Watchdog
Summary
Conclusion & Future Work
Long Distance Network (Rural areas)
Problems
• Travelling
to remote
node
•Unavailability of
trained personnel
•Simple solution :
Reboot the system
Ashwini Network Topology Image : [Bhaskar, 2007]
Self-healing and Self-management Capabilities required
Indian Institute of Technology, Bombay
Problem Statement
• Design and develop a centralized Network
Management & Monitoring system that
• helps in conducting experiments between any two
nodes in network from a central server.
• Detects, Diagnose and Resolve faults in the network
from a central location, else notifies admin about its
root cause.
Indian Institute of Technology, Bombay
Faults
Power Reset
H/W Quirk
Insufficient Tx Power
Link Misalignment
Interference
Packet Dropping at Receiver
MAC Level ACK Timeout
Network Partition
S/W or H/W Failure
S/W in Unknown State
Consequences: Low Throughput, High Packet Error Rate,
High Response Time
But, what is the Root Cause ?
Problem Statement
• Design and develop a centralized Network
Management & Monitoring system that
• helps in conducting experiments between any two
nodes in network from a central server.
• Detects, Diagnose and Resolve faults in the network
from a central location, else notifies admin about its
root cause.
• Gain practical experience: Establish Wi-Fi Mesh
Network within IIT Bombay Campus (Testbed).
Indian Institute of Technology, Bombay
Prior Work
1. Per packet Information at receiver in proc buffer.
[Bhaskar,2006].
Problems
Low level MAC Layer info
time,
seqno., and
length,
• Read information inArrival
string
format
wrote into a
type, RSSI ,Noise level,
file at runtime.
addresses etc.
• Lost lots of packets due to this slow processing.
Solution
Write & read from proc in Byte format (50 Byte
per packet) instead of string using dynamic
memory allocation.
Indian Institute of Technology, Bombay
Prior Work
2. Daemon running on nodes, ready to take instructions
form the server. [R. Shrivastava, 2006]
•
•
•
•
Problem
Memory Leaks
Consequences
Allocated dynamic
Daemon usually went into unknown
states.
memory during every
experiment
& never
freed
Caused wipe out of the Kernel
modules
(Madwifi).
it in ever lasting daemon
Even the OS including grub use to wipe out.
Stale file system.
Prior Work
2. Daemon running on nodes, ready to take instructions
form the server. [R. Shrivastava, 2006]
Problems
• Inappropriate design of error rate & throughput
experiments.
• Simultaneously read proc buffer at
packets at UDP and
slowed the
• Less flexibilityMAC
of and
experimental
parameter
values.
process.
• Wrong calculations
of thput
and error rates.
• Bug reported
in [Dheeraj,2007],
• Didreceived
not allowmore
the receiver
packets at
in Monitor
UDP thanmode.
at
•• Waited
for
synchronization
after
starting
timer
Receiver
MAC. in Monitor mode gives better results.
during the experiment.
• Removing log file of previous experiment took
considerable time during experiment.
Prior Work
2. Daemon running on nodes, ready to take instructions
form the server. [R. Shrivastava, 2006]
•
•
•
•
Other Problems
Used XML library which took virtually infinite time.
HardcodedConverted
addresses,
fileformat)
names and
array
sizes.
Central
server timed
out
XML file (text
resulting in experiment failure
to associative array.
Cases of system going into infinite loop.
Unnecessary time, space & calculations for things
which could have been done in simple manner.
(std. deviation, 0(n²) calc. )
Indian Institute of Technology, Bombay
Prior Work
3. Link Diagnosis, identify various faults on a long distance
link.[Dheeraj,2007]
Fault
Our Assessment
Power Reset
Correctly designed & implemented
Hardware Quirk
Incorrect design & implementation
Link Misalignment
Partially correct Implementation
Insufficient Tx Power
Partially correct design
Interference
Incomplete Detection
Packet Dropping at Receiver
Could not detect due to bug.
MAC level ACK
Incorrect design & Implementation
Shortcomings of Prior Work
1. Per packet Information at receiver in proc buffer.
[Bhaskar,2006].
2. Daemon running on nodes, ready to take
instructions form the server. [R. Shrivastava, 2006]
3. Link Diagnosis, identify various faults on a long
distance link.[Dheeraj,2007]
4. Existence of RSSI threshold beyond which error
rate is close to zero.[Bhaskar,2008]
Indian Institute of Technology, Bombay
Building over Prior Work
• Redesigned and re-implemented experiments
designed in [R.Shrivastava,2006] to use in link
diagnosis.
• Used & kept the basic implementation of daemon
almost intact.
• Redesigned most of the parts of link diagnosis
algorithm introduced in [Dheeraj,2007].
• Completely re-implemented the link diagnosis
algorithm.
Indian Institute of Technology, Bombay
Outline
•
•
•
•
Motivation & Problem Statement
Related Work
Prior Work and their shortcomings
Overall Approach
– Experiment Daemon & Types of experiments
– Link Diagnosis Algorithm
• IIT Bombay WiFi Mesh Network & Watchdog
• Evaluation and its Result
• Summary, Conclusion & Future Work
exptd
(Experiment daemon)
Performing Experiment
Mikrotik board
host1
sender
Gateway
Server
Indian Institute of Technology, Bombay
•
•
•
•
•
TCP Throughput
Packet Error Rate
UDP Throughput
Get Config
Set Config
host2
Receiver
Get Config
• Gets the current wireless configurations of the two
nodes in a link.
• Similarly, gets the default wireless configurations.
Wireless Configurations
IP Address
Channel
Essid
Mode
Transmission Power
Transmission Rate
• Gets RSSI history of the node (Moving avg. of avg.
RSSI of packet error rate experiments).
Indian Institute of Technology, Bombay
Set Config
• Sets the default wireless parameters of the
nodes on a link.
• Default parameters are stored in a file on
every node.
Indian Institute of Technology, Bombay
TCP Throughput
• Input :
Input Parameters
Packet Size
Wireless modes
Transmission Power
Duration
IP Addresses
Transmission Rate
• Output: TCP throughput between two nodes
in a link.
Indian Institute of Technology, Bombay
Packet Error Rate
Input Parameters
Packet Size
Wireless modes
Transmission Power
No. of packets to send
Inter packet gap
Transmission Rate
• Sender: sends packets through UDP socket.
• Low level information from proc for every pkt rcvd.
• Output: Packet error rate, # duplicate packets
For every MAC from whom packets received
Avg. RSSI
Max, Min RSSI
Avg. Noise
# of packets
# of packets with error
Inter packet gap
UDP Throughput
Input Parameters
Packet Size
Wireless modes
Transmission Power
Duration
Inter packet gap
Transmission Rate
Inter packet gap
• Madwifi driver sends packet as received by lower
layer to the transport layer.
Radiotap hdr MAC hdr
(26 B)
(32 B)
IP hdr
(20 B)
UDP hdr
(8 B)
DATA
CRC bytes
(4 B)
• Admin has option of receiver in Monitor Mode also.
• No reading of per packet information from proc.
• Output : UDP throughput, # packet received.
Link
Diagnosis
Inducing
Fault
Algorithm
exptd
(Experiment daemon)
Mikrotik board
Gateway
Faults
Power Reset
Hardware Quirk
Link Misalignment
Insufficient Tx Power
Interference
Packet Dropping
Server
MAC level ACK timeout
Start
Step 1
Read the command line
experiment parameters
Get the default & current
config of nodes (get_config)
Current txpower <
Default txpower
Y
N
Packet Error Rate exp. with
given parameters
Indian Institute of Technology, Bombay
POWER RESET
Set txpower to default
(set_config)
Power Reset
Cause
Power Outages and Voltage fluctuations at rural
areas.
Induction
Logged into a node using ssh and changed the tx
power using iwconfig command.
Indian Institute of Technology, Bombay
Start
Step 1
Read the command line
experiment parameters
Get the default & current
config of nodes (get_config)
Current txpower ≠
Default txpower
Y
N
Packet Error Rate exp. with
default parameters
Indian Institute of Technology, Bombay
POWER RESET
Set txpower to default
(set_config)
Hardware Quirk
• Packets with low transmit power sent at
regular interval.
WiFi Cards
Min RSSI Max RSSI
Avg. RSSI Avg.– Min
H/W Quirk
PER(%)
Ubiquiti SR2
-56
-38
-41
15
Yes
0.00
Ubiquiti XR2
-43
-38
-40
3
No
0.00
Note: Here, PER is the Packet Error Rate; RSSI in dBm
Indian Institute of Technology, Bombay
Start
Step 1
Read the command line
experiment parameters
Get the default & current
config of nodes (get_config)
Current txpower <
Default txpower
Y
N
Packet Error Rate exp.
Default parameters
Indian Institute of Technology, Bombay
POWER RESET
Set txpower to default
(set_config)
(Avg – Min) RSSI > 5
N
RSSI <
RSSI_HISTORY
N
Packet_err_rate
< 4%
Y
Possible Hardware Quirk
[Dheeraj,2007]
Y
Possible Link
Only Avg. RSSI of packets with
Misalignment
error compared with -75
Y
Link Working Well
Y
Txpower <
Max_Txpower
Step 2
N
Avg. SNR <
SNR_threshold
N
N
Insufficient Max Txpower
for link to operate
Report packet_err_rate for
lower tx-rates
Indian Institute of Technology, Bombay
Y
Packet Error Rate exp.
with Max txpower
Link Misalignment
• Observed in long distance links with
directional antennas.
• We get this fault often when we reduce the
default tx power or attach attenuators to the
nodes.
Indian Institute of Technology, Bombay
(Avg – Min) RSSI > 5
Y
Possible Hardware Quirk
N
RSSI <
RSSI_HISTORY
Y
Possible Link
Misalignment
N
Packet_err_rate
<4%
[Dheeraj,2007]
Y
Linkthe
Working Well
RSSI_HISTORY was
RSSI of last PER exp.
Step 2
N
Avg. SNR <
SNR_threshold
N
Y
Txpower <
Max_Txpower
N
Insufficient Max Txpower
for link to operate
Report packet_err_rate for
lower tx-rates
Indian Institute of Technology, Bombay
Y
Packet Error Rate exp.
with Max txpower
Insufficient Tx power
• Initially without any attenuator and Max tx
power, Packet error rate almost zero.
Indian Institute of Technology, Bombay
(Avg – Min) RSSI > 5
Y
Possible Hardware Quirk
N
RSSI <
RSSI_HISTORY
N
Y
Possible Link
Misalignment
Packet_err_rate
<4%
Y
Link Working Well
Y
Txpower <
Max_Txpower
Step 2
N
Avg. SNR <
SNR_threshold
N
N
Insufficient Max Txpower
for link to operate
Report packet_err_rate for
lower tx-rates
Indian Institute of Technology, Bombay
Y
Packet Error Rate exp.
with Max txpower
Insufficient Tx power
• Initially without any attenuator and Max tx power,
Packet error rate almost zero.
• With 36 dB attenuator and tx power as 8 dBm less
than the MAX (16 dBm) in lab in the absence of
interference on 802.11a at 54Mbps
Tx Power
(dBm)
Avg. RSSI
(dBm)
RSSI Threshold
at 54 Mbps
Packet error
rate (%)
Insufficient
Tx power
8
-85
-76.5
99.70
Yes
16
-78
-76.5
95.80
Yes
Indian Institute of Technology, Bombay
(Avg – Min) RSSI > 5
Y
Possible Hardware Quirk
N
RSSI <
RSSI_HISTORY
N
Y
Possible Link
Misalignment
Packet_err_rate
<4%
Y
Link Working Well
Y
Txpower <
Max_Txpower
Step 2
N
Avg. RSSI <
RSSI_threshold
N
Packet Error Rate exp.
with Max txpower
[Dheeraj,2007]
N
Did not gave any such
Insufficient Max Txpower information to the admin.
for link to operate
Report packet_err_rate for
lower tx-rates
Indian Institute of Technology, Bombay
Y
Interference
• Interference already present on the KReSIT-SOM
link.
• Experimented with tx power as 1 dBm and 10 dB
attenuator, then with MAX tx power at 54Mbps.
• Found Capture effect.
Tx power (dBm)
Avg RSSI (dBm)
Packet error rate (%)
1
-69
95.80
16
-52
9.70
Indian Institute of Technology, Bombay
Txpower <
MAX
Y
Perform PER exp.
with MAX tx power
Improved
PER
Y
N
N
Capture Effect ,
Insufficient tx
power, Interference
Check MAC level per packet
info for other Wi-Fi sources
[Dheeraj,2007]
• Did not check whether the
result is improved in case of
N
Other Wi-Fi
N
MAX Tx power
Avg. Noise > -92
sources
• So, no capture effect and
insufficient Tx power was
Y
Y
pointed out.
INTERFERENCE from other Wi-Fi sources
INTERFERENCE from non Wi-Fi sources
Report PER for lower Tx-Rates
Report PER for other channels
Step 2
Txpower <
MAX
Y
Perform PER exp.
with MAX tx power
Improved
PER
Y
N
N
Capture Effect &
Interference
Check MAC level per packet
info for other Wi-Fi sources
Other Wi-Fi
sources
N
Avg. Noise > -92
Y
Y
INTERFERENCE from other Wi-Fi sources
INTERFERENCE from non Wi-Fi sources
Report PER for lower Tx-Rates
Report PER for other channels
Step 2
N
Packet Dropping at Receiver Kernel
Low processing speed of system, causes packet dropping
at transport layer
• Sent 100 Bytes UDP payload packets at 54 Mbps.
• Successfully detected packet dropping at receiver
kernel.
Step 2
UDP throughput Experiment
N
UDP thput
< 80% of
expected
Y
Packet Error Rate exp. with same
number of packets as in UDP
# packets
received at
MAC > UDP
N
Y
PACKET DROPPING at
receiver kernel
Report UDP thputs achieved at
different inter-packet gaps
Step 3
Indian Institute of Technology, Bombay
Increase inter-packet
gap by 0.25 ms
Step 2
UDP throughput Experiment
N
UDP thput
< 80% of
expected
[Dheeraj,2007]
Could not detect this,
because of the bug reported.
Y
Packet Error Rate exp. with same
number of packets as in UDP
# packets
received at
MAC > UDP
N
Y
PACKET DROPPING at
receiver kernel
Report UDP thputs achieved at
different inter-packet gaps
Step 3
Indian Institute of Technology, Bombay
Increase inter-packet
gap by 0.25 ms
Step 2
UDP throughput Experiment
N
UDP thput
< 80% of
expected
Y
Packet Error Rate exp. with same
number of packets as in UDP
# packets
received at
MAC > UDP
N
Y
PACKET DROPPING at
receiver kernel
Report UDP thputs achieved at
different inter-packet gaps
Step 3
Indian Institute of Technology, Bombay
Increase inter-packet
gap by 0.25 ms
Packet Dropping at Receiver Kernel
Low processing speed of system, causes packet dropping
at transport layer
• Sent 100 Bytes UDP payload packets at 54 Mbps.
• Successfully detected packet dropping at receiver
kernel.
Inter pkt # Packets # Rcvd at # Rcvd at
gap (µsec)
Sent
MAC
UDP
MAC-UDP
Packet UDP thput PER (%)
(%)
Dropping
(Mbps)
172
34066
28832
20249
29.77
Yes
3.238
15.36
201
25000
24825
22245
10.39
Yes
3.558
0.70
301
16667
16565
16321
1.47
Yes
2.611
0.61
402
12500
12406
12342
0.52
No
1.975
0.75
MAC level ACK Timeout
• Observed only on long distance links.
Indian Institute of Technology, Bombay
Step 3
Packet error rate exp. with both
nodes in adhoc mode
High no. of
duplicates
Y
Possibility of MAC level ACK
timeout
END
Indian Institute of Technology, Bombay
N
Step 3
Packet error rate exp. with both
nodes in adhoc mode
[Dheeraj,2007]
• Performed the TCP throughput
experiment
with & without MAC
High no. of
N
ACK.
duplicates
• Goes completely wrong in
presence of interference.
Y
• At 54 Mbps & MAX tx power on
KReSit-SOM link we observed :
Possibility of MAC level ACK with MAC ACK
: 21.53 Mbps
timeout
without MAC ACK : 4.27 Mbps
END
Indian Institute of Technology, Bombay
Indian Institute of Technology, Bombay
Wi-Fi Nodes in Campus
• 8 dBi Omni-directional
Antenna.
• Soekris board with Atheros
802.11 b/g cards.
• Madwifi driver with 27dbm
max transmit power.
• Soekris powered using
Ethernet cable and Power
Over Ethernet (P.O.E.)
adapter.
•Watchdog and exptd running
continuously.
Antenna and water proof box containing soekris at KRESIT roof
Watchdog
• Reset of wireless parameters of a node causes:
• Higher packet error rate.
• Node disconnected from rest of the network.
• User program in unknown state or not running.
• Stale file system especially on Soekris boards.
• Hence, a need of watchdog to keep the node in
known state and connected to rest of the network.
Indian Institute of Technology, Bombay
Working of Watchdog
START
NFS stale?
Replace using
repository
Yes
No
No
Exptd Running?
Time > 200 min
since last exp?
Yes
No
Reboot
Start Exptd
Yes
No
Exptd in good state?
No
Yes
Time > 10 min
Since last exp?
Yes
Kill Exptd
Restart Exptd
Start Exptd
Wait for 5 min
Mikrotik Vs Soekris Boards
• Soekris board CPU speed
133 MHz (net 4521) & 266 MHz (net 4826)
• Mikrotik RB 433 AH processing speed : 680 MHz
Indian Institute of Technology, Bombay
UDP Throughput (Mikrotik Vs Soekris)
Tx rate
(Mbps)
UDP Thput (Mbps)
Soekris (266 MHz)
UDP Thput (Mbps) Expected UDP
Mikrotik (640 MHz) thput (Mbps)
Inter packet
gap (ms)
6
5.4
5.52
5.65
2.08
9
7.92
8.16
8.39
1.4
12
10.33
10.80
11.09
1.06
18
15.24
15.86
16.35
0.72
24
12.02
20.68
21.43
0.55
36
11.57
29.85
31.08
0.38
48
11.83
37.16
40.11
0.29
54
11.70
41.73
44.42
0.26
802.11a, No backoff
Waits for SIFS = 17.5 µsec time before sending next packet
Indian Institute of Technology, Bombay
Monitor mode Vs Adhoc mode
Tx Rate
(Mbps)
UDP throughput
Ahdemo (Mbps)
UDP throughput
Monitor (Mbps)
Expected UDP
throughput(Mbps)
Inter packet
Gap (µsec)
6
5.4
5.4
5.65
2.08
9
7.9
7.92
8.39
1.4
12
8.8
10.33
11.09
1.06
18
5.2
15.24
16.35
0.72
24
2.1
12.02
21.43
0.55
36
1.7
11.57
31.08
0.38
48
1.9
11.83
40.11
0.29
54
1.7
11.70
44.42
0.26
For Soekris Board at 802.11a in lab
Indian Institute of Technology, Bombay
Monitor mode Vs Adhoc mode
Tx Rate
(Mbps)
UDP throughput
Ahdemo (Mbps)
UDP throughput
Monitor (Mbps)
Expected UDP
throughput(Mbps)
Inter packet
Gap (µsec)
6
5.45
5.52
5.65
2.08
9
8.09
8.16
8.39
1.4
12
10.73
10.80
11.09
1.06
18
15.79
15.86
16.35
0.72
24
20.61
20.68
21.43
0.55
36
29.73
29.85
31.08
0.38
48
37.02
37.16
40.11
0.29
54
37.34
41.73
44.42
0.26
For Mikrotik Board at 802.11a in lab
Indian Institute of Technology, Bombay
Summary of Work Done
• Redesigned and re-implemented experiments
keeping the exptd daemon almost intact.
• Changed Madwifi driver code:
1. Improved process of getting per packet info from proc.
2. ACK disable at runtime during the experiment.
3. MAC backoff disable, wait only for SIFS before sending.
• Redesigned parts of link diagnosis algorithm.
Indian Institute of Technology, Bombay
Summary of Work Done
• Re-implemented the link diagnosis algorithm
with the new design.
• Designed and implemented a Watchdog for
nodes in the network.
• Established a WiFi mesh network within IIT
Bombay.
Indian Institute of Technology, Bombay
Conclusion
• We are able to successfully detect, diagnose and
even rectify faults common in long distance WiFi
network at rural areas from a central location.
• We gained practical experience by establishing
WiFi mesh network within the IIT Bombay
campus.
• We can keep every node in a network in known
state and connected to rest of the network with
the help of watchdog.
Indian Institute of Technology, Bombay
Future Work
• System can be extended to work for multi hop links.
• Currently end nodes assumed to be wireless. Can
be extended to end nodes being wired.
• A user friendly GUI required at the central server.
Indian Institute of Technology, Bombay
Thank You!
Questions?
Indian Institute of Technology, Bombay
Related Work
• Offline Diagnosis : Collect traces from nodes, merge
then infer faults. [R. Mahajan, 2006]
• Use Simulator : Collect data from network, feed to
simulator, predict using difference in observed and
expected performance. [Lili Qiu, 2006]
• VSAT & GPRS backlinks, S/w & h/w watchdog. Uses
s/w & h/w redundancy. [Surana, 2008]
None of the above work handle all the faults
addressed in our work.
Indian Institute of Technology, Bombay
Evaluation
• Power Reset : Manually change the tx power by
logging into a node. (Tested)
• Hardware Quirk : Depends on cards been used.
(Tested)
• Link Misalignment : Can be tested on long distance
link with directional antennas.
• Insufficient Tx Power : Attach attenuators to a
node & reducing the tx power value. (Tested)
Indian Institute of Technology, Bombay
Evaluation (contd…)
• Interference : Found almost every where in campus
for 802.11 b/g. (Tested)
• Packet dropping at receiver kernel : Send small 100
Bytes packets with 54 Mbps. (Tested)
• MAC level ACK Timeout : Observed on long
distance link only.
Indian Institute of Technology, Bombay
Experimental Setup
• KReSIT-SOM Link
• 15 dbi omni directional antennas at both
ends.
• Working on 802.11 g.
• Ubiquiti XR2 Wifi cards
Indian Institute of Technology, Bombay
Experimental Results
Indian Institute of Technology, Bombay
Ubiquiti XR2 Cards
Tx -power
Min RSSI
Max RSSI
Avg. RSSI
Avg. - Min H/w Quirk
PER
1 dbm
-61 dbm
-55 dbm
-57 dbm
4 dbm
No
11.90 %
2
-57
-51
-53
4
No
4.60
3
-57
-51
-53
4
No
6.20
4
-55
-50
-52
3
No
4.20
5
-55
-51
-52
3
No
0.40
6
-53
-48
-51
2
No
0.30
7
-50
-46
-48
2
No
0.20
8
-51
-46
-48
3
No
0.70
9
-50
-45
-48
2
No
0.60
10
-49
-44
-46
3
No
0.10
11
-48
-44
-45
3
No
0.00
12
-48
-43
-44
4
No
0.00
13
-45
-40
-42
3
No
0.10
14
-45
-40
-42
3
No
0.40
15
-45
-40
-42
3
No
0.20
16
-43
-38
-40
3
No
0.00
Insufficient Txpower
Tx Rate
(Mbps)
Tx power = 1dbm
Tx power = 16 dbm (Max)
RSSI (dbm)
PER (%)
RSSI (dbm)
PER (%)
6
-66
89.60
-50
1.20
9
-66
94.00
-50
1.80
12
-67
94.90
-49
1.40
18
-67
88.60
-49
1.30
24
-67
90.70
-49
2.00
36
-69
99.70
-50
1.60
48
-68
94.20
-51
2.90
54
-69
95.80
-52
9.70
10 db Attenuator, channel 8
Indian Institute of Technology, Bombay
10 dB Attenuators, channel 8 (Interference present)
Tx power (dbm)
Avg. RSSi (dbm)
PER (%)
1
-66
89.80
2
-65
81.00
3
-62
78.60
4
-63
81.80
5
-61
62.90
6
-60
46.20
7
-61
36.20
8
-59
29.20
9
-56
25.90
10
-55
26.40
11
-53
18.50
12
-52
23.10
13
-52
5.30
14
-51
1.60
15
-51
0.30
16
-50
0.10
Packet Dropping at Receiver Kernel
Inter pkt # Packets # Rcvd at # Rcvd at
gap (µsec)
Sent
MAC
UDP
MAC-UDP
Packet UDP thput PER (%)
(%)
Dropping
(Mbps)
172
34066
28832
20249
29.77
Yes
3.238
15.36
201
25000
24825
22245
10.39
Yes
3.558
0.70
301
16667
16565
16321
1.47
Yes
2.611
0.61
402
12500
12406
12342
0.52
No
1.975
0.75
502
10000
9945
9935
0.10
No
1.590
0.55
602
8334
8294
8284
0.12
No
1.324
0.48
UDP Payload = 100 Bytes , Tx rate = 54 Mbps
Indian Institute of Technology, Bombay