TCP Tahoe, Reno and Vegas

1. MINISTRY OF EDUCATION
THE NINTH NATIONAL CONFERENCE ON
SCIENCE AND ENGINEERING, 2016
UPPER MYANMAR
8-Jul-2016
Supervised by
Dr. May Zin Oo
Presented by
Ma Chit Su Khin
ME-IT-2(13th batch)

2. Outline of Presentation
2
Aim and Objectives
Introduction
Research Methodology
Simulation Parameters
Simulation Scenarios
Simulation Results
Discussion and Conclusion
References

3. Aim and Objectives
Aim:
To analyze the performance of the most widely used TCP variants
(Tahoe, Reno and Vegas) in MANET for ensuring which variant
performs well under the different scenarios
Objectives:
Learning theories of TCP variants: TCP Tahoe, Reno and Vegas
Designing the MANETs model using INET framework in OMNeT++
simulator
Implementing the TCP setting in these MANETs including TCP
variants
Evaluating the performances of those TCP variants and compare the
results
Identifying the most suitable TCP variant in MANETs with different
network scenarios
3

4. Introduction
Introduction: A Mobile ad hoc network (MANET)
Collection of two or more devices
A network where the node can communicate freely
Communication of mobile nodes without requiring any base station
Usefulness in disaster recovery situations and the places with non-
existing or damaged communication infrastructure
4

5. Introduction: Transmission Control Protocol (TCP)
A very important role to transfer data packets efficiently and reliably
A transport layer protocol that is connection oriented protocol
A reliable and in sequence delivery of data between two nodes
Introduction: TCP over MANET
Reliability is important in the applications of MANET.
TCP can provide the reliability requirements of MANET applications.
TCP performance is not satisfactory as wired networks.
Introduction
5

6. Introduction: Problem Statements
TCP performance is not always stable, changing depending on
environments.
Due to scalability and mobility in MANETs, TCP encounters
several performance problems.
It is important to choose the most suitable TCP variant under
the specific network scenarios.
Introduction
6

7. Research Methodology (RM)
Construct a MANET and implement the most widely used TCP
variants (TCP Tahoe, Reno and Vegas) on those MANET
Use OMNeT++ simulator together with INET framework to add
TCP protocols and to set the simulation scenarios
Analyse the performance of TCP variants on OMNeT++ with
different simulation parameters by means of measuring
throughput and delay
7

8. RM: TCP Variants
The number of TCP variants are developed to improve the
performance of TCP congestion control algorithm.
The most widely used TCP variants are:
I TCP Tahoe
I TCP Reno
I TCP new Reno
I TCP Vegas
I TCP Westwood
I ...
8

9. RM: TCP Tahoe
Starts the slow start algorithm when the connection is established.
I Set congestion window (cwnd ) = 1 MSS.
I slow start threshold (ssthresh) = pre-agreed value (normally a
multiple of maximum segment size: MSS).
Increase cwnd by one segment if each of the acknowledgement (ACK)
arrives (Double cwnd for each RTT) before the retransmission timer
expires.
Continue increasing the window exponentially until it reaches
ssthresh .
Increases cwnd by one segment for each round trip time (RTT) once
the congestion window reaches ssthresh .
9

10. Slow startStart
None
ssthresh = …
cwnd = 1
Time-out or 3 dupACKs
ssthresh = cwnd/2
cwnd = 1
An ACK arrived
cwnd = cwnd + 1
Time-out or 3 dupACKs
ssthresh = cwnd/2
cwnd = 1
Congestion
avoidance
An ACK arrived
cwnd = cwnd + (1/cwnd)
cwnd ≥ ssthresh
None
RM: TCP Tahoe
1
0

11. RM: TCP Reno
If three duplicate ACKs arrive, TCP moves to the fast-recovery state
and remains there as long as more duplicate ACKs arrive.
I cwnd grows exponentially, but starts cwnd = ssthresh + 3 MSS
When TCP enters the fast-recovery state, three major events may
occur.
I If duplicate ACKs continue to arrive, TCP stays in this state, but the
cwnd grows exponentially.
I If a time-out occurs, TCP assumes that there is real congestion in
the network and moves to the slow start state.
I If a new ACK (non-duplicate) arrives, TCP moves to the congestion
avoidance state.
11

12. 12
RM: TCP Reno
12

13. RM: TCP Vegas
TCP Vegas modifies TCP Reno in the slow start, congestion avoidance,
and retransmission mechanism.
New Congestion Avoidance algorithm
I Diff = Expected − Actual Throughput
I Actual Throughput =
I Expected Throughput =
cwnd
BaseRTT
I The BaseRTT of a TCP connection is the RTT of the connection
when it is not congested
I BaseRTT = min ( RTT of packets observed )
13
Bytes sent in SampleRTT
SampleRTT

14. RM: TCP Vegas
I TCP Vegas defines 2 throughput threshold values: α < β (Typically
1, 3 or 2, 4 pairs of packets)
I If (Diff < α), increase cwnd linearly
I If (Diff > β), decrease cwnd linearly
I If (α ≤ Diff ≤ β), cwnd is unchanged
TCP Vegas Slow Start:
I TCP Vegas increases the cwnd exponentially only every other RTT.
I When the actual rate falls below the expected rate, TCP Vegas
changes from slow start mode to linear increase/decrease mode.
14

15. RM: TCP Vegas
TCP Vegas retransmission:
I TCP Vegas uses a more accurate timer.
I It can determine the packet loss event from one duplicate
ACK.
I When a duplicate ACK is received, TCP Vegas checks
whether the retransmission is performed or not.
I If ((Now − Packet Transmission Time) > Timeout for
packet), then TCP Vegas will retransmit the packet
immediately.
15

16. RM: Network Topology
16

17. RM: Internal Structure of the Host
17

18. Attributes Values
Constraint Area 600 m × 400 m
Simulation Time 300 s
Bitrate 2 Mbps
Advertised Window Size 65535
Maximum Segment Size (MSS) 1500
Maximum transfer unit 2000 B
Send Bytes 1 MB
Mobility speeds 20 mps, 10 mps, 5 mps
Number of nodes 10, 20, 30, 40 nodes
Simulation Parameters
18

19. Simulation Scenarios (SS)
File Transfer Protocol (FTP) is used in application layer.
The application type of sender site is TCPSessionApp and receiver
site is TCPSinkApp.
The throughput of TCP variants in MANET are evaluated and
compared.
Throughput of the whole network can be defined by:
The performance analysis are done on OMNeT++ with different
simulation parameters:
I Number of nodes
I Mobility speeds
19
Node Throughput of Data Transmission
Total Number of Nodes

20. SS: Network Animation
20

21. 0
2000
4000
6000
8000
10000
12000
10 nodes 20 nodes 30 nodes 40 nodes
Throughput(bps)
Number of nodes
Mobility speed = 5 mps
Tahoe
reno
vegas
Figure. Throughput of TCP variants at 5 mps mobility speed
Simulation Results
21

22. 0
1000
2000
3000
4000
5000
6000
7000
10 nodes 20 nodes 30 nodes 40 nodes
Throughput(bps)
Number of nodes
Mobility speed = 10 mps
Tahoe
Reno
Vegas
Figure. Throughput of TCP variants at 10 mps mobility speed
Simulation Results
22

23. 0
2000
4000
6000
8000
10000
12000
14000
10 nodes 20 nodes 30 nodes 40 nodes
Throughput(bps)
Number of nodes
Mobiity speed = 20 mps
tahoe
reno
vegas
Figure. Throughput of TCP variants at 20 mps mobility speed
Simulation Results
23

24. Discussion and Conclusion
The throughput of the commonly used TCP variants (Tahoe,
Reno, Vegas) are evaluated in MANET.
At 5 mps,
I TCP Vegas performs almost 7% better than Reno and 14%
better than Tahoe for the 10 nodes network size.
I Vegas gives a better performance in all network sizes.
I The throughput of Tahoe and Reno is comparable at 5 mps.
At 10 mps,
I Vegas is almost 12% higher than Reno and almost 22%
higher than Tahoe.
I Reno is slightly better than Tahoe at 10 mps node speed.
24

25. At 20 mps,
I The throughput of Reno is almost 18% higher than Tahoe and 6%
lower than Vegas in 10 nodes network sizes.
I Vegas performs almost 24% higher than Tahoe.
For all network sizes and different mobility speeds, Vegas performs
better than TCP Tahoe and TCP Reno for all simulation scenarios.
The overall throughput of TCP Reno is slightly higher than that of TCP
Tahoe.
25
Discussion and Conclusion

26. References
26
1) Akinwole A.K. and Emuoyibofarhe O.J, "Performance Comparison of Mobile Ad Hoc
Network Routing Protocol and Transmission Control Protocol", ISSN 1913-8989 E-
ISSN 1913-8997, Vol. 5, No. 4; 2012.
2) Dr. Madhu Goel and Ms. Divya Garg, "Performance Evaluation of TCP New Reno and
TCP Westwood Over AODV in Mobile Ad-Hoc Network" IJRTS,ISSN (online): 2348-
1439 , Vol. 1, Issue 7, June 2014 .
3) Debajyoti Pal and Gopal Kundu, "Performance Analysis of Different MANET Specific
TCP variants", IJECSE, Volume1, Number 4.
4) Ravinder Kaur and Gurpreet Singh Josan, "Performance Evaluation Of Congestion
Control Tcp Variants In Vanet Using Omnet++", (IJERA) ISSN: 2248-9622
www.ijera.com Vol. 2, Issue 5, September- October 2012.
5) Behrouz A. Forouzan ,"Data Communications AND Networking", 5th ed, McGraw-
Hill, a business unit of The McGraw-Hill Companies, Inc., 1221 Avenue of the
Americas, New York, NY 10020 , 2012.
6) May Zin Oo and Mazliza Othman, "Analytical Studies of Interaction between
Mobility Models and Single-Multi Paths Routing Protocolsin Mobile Ad Hoc
Networks", Wireless Pers Commun DOI 10.1007/s11277-010-0205-3 © Springer
Science+Business Media, LLC. 2010.

28. 28
Figure 1. Illustration of slow start algorithm

29. congestion avoidance algorithm

30. Fast retransmission algorithm