Transcript An Integrated Approach for Performance Modeling and

Performance Evaluation of
TCP Throughput on Wireless
Cellular Networks
Osaka University
Masahiro Miyoshi, Masashi Sugano,
Masayuki Murata
[email protected]
Contents
Backgrounds
 Objectives
 Network model
 Analysis method
 Evaluation
 Conclusion

Backgrounds
IMT-2000 Network in JAPAN
Collision detection
( Using Slotted ALOHA )
TCP
Mobile
Terminal
Mobile
Terminal
Base
Station
TCP
Transmission error recovery
( Using FEC,ARQ )
Past researches
Evaluate TCP performance
on Wireless channels
Influence of transmission error
 Influence of handoff

No consideration influence of lower
layer protocols above researches
Objectives
Evaluate TCP performance

By considering collision detect protocol
Slotted ALOHA

By error correct protocol
ARQ, FEC
Network Model
Wireless Link
Wired Link
Mobile
Terminal
Base
Station
TCP Source
TCP Source
・・
TCP Source
Mobile
Switching
Center
Bottleneck
Link
Destination
packet loss in bottleneck buffer
packet loss by transmission error
packet loss by collision
Network Model
Our View
about packet loss on TCP
・packet loss in bottleneck buffer
・packet loss by transmission error
→ packet loss on TCP layer
・packet loss by collision
(using Stop and Wait retransmission)
→ delay on TCP layer
Analysis Method

Slotted ALOHA throughput
S ALOHA  G exp( G)
N: packet retransmission interval
L: packet length
Terminal A
Terminal B

Slotted ALOHA delay

×
collision
DALOHA   (i  1) NL(1  S ALOHA)i S ALOHA
i 0
Analysis Method

TCP throughput
Packet loss rate ←Parameter
Round trip time ← DALOHA
STCP 
1
2bp
3bp
2
RTT
 To min( 1,3
) p(1  32 p )
3
8
Time out time ← DALOHA
Parameter sets
0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
Bandwidth of wireless link : 2Mbps
Bandwidth of wired link : 125Kbps
The number of mobile Terminal: 30nodes
TCPsegment size: 100Bytes
packet loss rate p
Evaluation
Improving throughput at the Slotted ALOHA does not
necessarily lead to the
TCP
throughput
improvement
1
10
100
1000
Retransmission Interval N
(packets)
TCP throughput
Slotted ALOHA throughput
40
490
Throughput (Kbps)
Throughput (%)
35
30
25
20
15
10
5
470
450
430
410
390
370
350
0
1
10
100
Retransmission Interval (packets)
1000
1
10
100
Retransmission Interval (packets)
1000
Parameter sets
Bandwidth of wireless link : 2Mbps
Bandwidth of wired link : 125Kbps
The number of mobile Terminal: 5nodes
TCPsegment size: 100Bytes
ARQ：Go back N, retransmission times 1
FEC(5bit correct)：Reed Solomon(127,117)
FEC(20bit correct)：Reed Solomon(127,87)
Evaluation
The use of FEC is effective
to prevent TCP throughput degradation
Throughput (Kbps)
TCP throughput
18
16
14
12
10
8
6
4
2
0
without ARQ FEC
ARQ
FEC ( 5bit correct )
FEC ( 20bit correct )
0
5
10
Noize level Eb/N0 (dB)
15
Conclusion
We have shown TCP performance
considering influence of lower layer protocols

Improving throughput at the Slotted
ALOHA does not necessarily lead to the
TCP throughput improvement

The use of FEC is effective to prevent
TCP throughput degradation