Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 22

TCP Cross-Layer Implementation -

Scheme over IEEE 802.11

March 12,2018

Prepared by: Guided By:


Vekariya Earnest Dr. Purvang Dalal
EC80 (14ECUOG052)
TCP Mechanisms

TCP has to ensure Reliable & Efficient transport.

Reliability using checksum, acknowledgements, Retransmissions


Efficiency by fair utilization of network resources.

TCP has two basic method for Congestion Control


Flow control
Slow Start
Congestion Avoidance
Loss Detection & Recovery
Fast Transmit
Fast Recovery
Flow Control
Slow Start
The slow start algorithm is based on the idea that the size of
congestion window (cwnd) starts with one maximum segment
size (MSS).
The size of the cwnd increases by 1 MSS on arrival of each
acknowledgement.
Slow start algorithm grows cwnd exponentially per RTT.

Congestion Avoidance
The size of the cwnd increases by 1/cwnd MSS on arrival of each
acknowledgement.
Congestion avoidance algorithm grows cwnd linearly (by 1 MSS)
per RTT.
TCP Mechanism Graph
Loss Recovery
Fast Retransmit
If three or more dupack are received in a row , it is
strong indication that the segment has been lost. TCP
then performs retransmission of what appears to be
the missing segment, without waiting for
retransmission timer to expire. This retransmission is
called fast retransmission.
Fast Recovery
After retransmitting the lost packet , TCP undergoes
Fast Recovery for solving problem of congestion in
the network.
Loss Recovery(cont.)

◦ When the third duplicate ack is arrived in a row is


received, set ssthresh to one half the current cwnd .
Retransmit the missing segment. Set cwnd to ssthresh
plus 3 times the segment size. This inflates the
congestion window by the number of segment that
has left the network.
Impact of RTT variation
 TCP decides the sending rate based on RTT.
 throghput = cwnd / RTT

Increase in RTT, causes slow convergence of cwnd and it


will have adverse impact on the throughput.

 RTT variations are believed as congestion indicator.


 If RTT variation is weakly correlated to the network
congestion, it leads to unnecessary impact on
throughput.
i.e. decrease the bandwidth utilization
Problem Definition
 TCP is mainly designed for wired network with an objective
to utilize maximum of available bandwidth without network
overloading.
 Development of technology and invention of portable device
have led to increase in use of wireless network.
 TCP started facing problem in wireless network such as
channel fading, mobility of node, route failure, etc. These
factors cause RTT variations despite there is no change in
the congestion.
 TCP revises sending rate based on the RTT of the network.
 When TCP detects the false (due to non-congestion
reasons) variation in RTT, it unnecessary reduces the
sending rate of the network.
Problem Definition(cont)
 Due to reduction in sending rate of TCP, the overall
throughput of TCP will be reduced.
 From the above mention reason, it can be derived that TCP
can not utilize the maximum available bandwidth particularly
when RTT variations are not inline with network congestion.

RTT variations due to reasons other than network congestion is inappropriate

This work is based on the idea to correct RTT variations if


influenced by the non-congestion delay components.
This work shows implementation of Cross-Layer Scheme
between TCP and IEEE 802.11 link layer for RTT correction.
Reason for variation in RTT
 Mobility of node
 Link Failure and Reestablishment
 High Bit Error Rate
Discussion on Approaches
 To solve the problem , it is required to correct RTT to
improve the efficiency.
 For improving TCP efficiency, it is required to calculate
retransmission delay and contention delay for predicting
the correct RTT.
 In our project, it is required to implement the cross
layer between transport layer and data link layer
Cross layer
Cross layer scheme is used to communicate between
two layers by the violation of OSI layer.
Logic Flow of mac_802.11
Logic flow for calculation of
contention delay
Logic flow for calculation of
propagation delay
Simulation results
 In our project, we considered the below topology for
the simulation of our project. In the given topology,
there is one wired link between s0 and n0 and there is
one wireless link between n0 and r0.
Simulation Result(cont.)
 Simulation Result of correct RTT
Error Round trip Propagation Contention Total delay Correct RTT Percentage
time delay delay change

0% 116.07 0.000443 0.177089 0.177532 115.89 0.115%

1% 38.72 0.148290 0.340071 0.488361 38.231 1.26%

5% 29.14 0.322870 0.502456 0.825327 28.314 2.834%

10% 28.56 0.366995 0.565596 0.932591 27.627 3.226%


Simulation Result(cont.)
 Comparison of throughput of original and correct RTT
Error RTT Correct RTT CWND Throughput1 Throughput2
=(cwnd/rtt) =(cwnd/correctrtt)

0% 116.07 115.89 43.72 0.376 0.3771

1% 38.72 38.231 12.25 0.316 0.320

5% 29.14 28.314 6.66 0.228 0.235

10% 28.56 27.627 6.15 0.215 0.222


Simulation Result(cont.)
 Improvement of throughput efficiency

Error Improvement
0% 0.2925%
1% 1.26%
5% 3.07%
10% 3.255%
Conclusion
 By performing above task, the following can be inferred
◦ The simulation was performed for both wired and
wireless topology and efficiency. From the simulation
result , it was concluded that efficiency reduce in
wireless network.
◦ Cross layer scheme is required for the correction of
RTT. For the correction of RTT, it is required to
calculate retransmission and contention delay and this
information can be transferred from data link layer to
transport layer using cross layer scheme.
Conclusion(cont.)

◦ The information obtained from the cross layer can be used for
the improvement of throughput. From these information, it is
possible to enable the fast convergence of congestion window
and improvement of overall throughput. It was concluded, it is
possible to improve the TCP efficiency by 2-3%.
Thank You

You might also like