www.ijcait.

com International Journal of Computer Applications & Information Technology

Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

Survey of Performance based Transmission Control

Protocol in MANET
Sapna Bagde1
Department of Information Technology
Barkatullah University Institute of Technology
Bhopal, M.P., India.
Prof. Poonam Sinha
Department of Information Technology
Barkatullah University Institute of Technology
Bhopal, (M.P.), India
Asst. Prof. Ashish Jain
Department of Information Technology
Barkatullah University Institute of Technology
Bhopal, (M.P.), India

Abstract: Transmission Control Protocol (TCP) is a networks and because of that its performance is quite
connection-oriented transport service that ensures the poor when it is required to work in the typically loss
reliability of message delivery. It verifies that messages wireless environment. Unlike cellular networks, where
and data were received. TCP provides reliable, ordered only the last hop is based on a wireless medium, ad hoc
delivery of a stream of bytes from a program on one networks are composed exclusively of wireless links,
computer to another program on another where multihop connections may be in place. Besides,
computer. TCP provides a communication service at an in an ad hoc scenario all nodes can move freely and
intermediate level between an application programs. unpredictably, which makes the TCP congestion control
TCP is the protocol used by major Internet applications quite hard since it is a clock based mechanism.
such as the World Wide Web, email, remote Consequently, the error-detection and error-recovery
administration and file transfer. TCP is a reliable strategies inherent in standard TCP need to be adapted
transport protocol that is well tuned to perform well in in order to fit this environment. In particular, since the
traditional networks. However, several experiments errors in this environment occurs not only due to
and analysis have shown that this protocol is not congestion but also due to medium constraints and
suitable for bulk data transfer in high bandwidth, large mobility, TCP needs to distinguish the nature of the
round trip time networks because of its slow start and error so that it can take the most appropriate action for
conservative congestion control mechanism. In this each case. Additionally, the emerging link and network
paper we discussed a survey of Performance Based layer algorithms for this kind of network can play a key
Transmission Control Protocol in Mobile Ad-hoc role on TCP performance. Likewise, factors such as
Network environment. The performance based path asymmetry (that may also be caused by lower
techniques are categorized based upon different layers strategies, among other elements) and congestion
approaches like throughput, end-to-end delay, window size might also impair the performance of this
congestion control etc. We also analysis the major protocol. Although there are a number of differences
improvement in recent methods for performance based between cellular and ad hoc networks, some of the ideas
TCP in MANET. developed in some proposed solutions for the former
can be used in the latter as well. In ad hoc environment
Keywords: MANET, TCP Techniques, Performances TCP can be led to persist mode whenever a long
Analysis. disconnection occurs and under wireless medium
induced losses it can simply retransmit the lost packet
Introduction instead of invoking its congestion control mechanism.
Both ideas have already been evaluated for cellular
Ad hoc networks are complex distributed systems that networks where only the last hop is of concern. Ad hoc
consist of wireless mobile or static nodes that can freely networks pose some tough challenges to TCP due to the
and dynamically self-organize. In this way they form fact that it was not designed to operate in such a highly
arbitrary and temporary ad-hoc network topologies, dynamic scenario in terms of topology. In reality, even
allowing devices to seamlessly interconnect in areas though TCP has evolved significantly over the years
with no pre-existing infrastructure. Depending on toward a robust and reliable service protocol, the focus
whether the wireless nodes are mobile or static, an ad has been primarily on wired networks. In this scenario,
hoc network is called a Mobile Ad hoc Network the additive-increase/multiplicative decrease strategies
(MANET). TCP was designed to work in wired

Page |9
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

coupled with the fast recovery and fast retransmit Transmission Control Protocol Working Phases:
mechanisms inherent in most of current TCP versions,
TCP protocol operations may be divided into three
provide an effective congestion control at the end nodes
phases. Connections must be properly established in a
[1-6]. In recent years, several schemes have been
multi-step handshake process (connection
proposed for satellite as well as for cellular wireless
establishment) before entering the data transfer phase.
networks, but much has still to be done on mobile ad
After data transmission is completed, the connection
hoc framework. In this sense, it is quite important the
termination closes established virtual circuits and
knowledge of both the main factors affection TCP
releases all allocated resources.
performance in such an environment and the actual
potentiality of the recent existing proposed solutions, A TCP connection is managed by an operating system
which are addressed in this paper. through a programming interface that represents the
local end-point for communications, the Internet socket.
Background on TCP During the lifetime of a TCP connection it undergoes a
series of state changes:
TCP techniques adjust the network congestion
avoidance parameters of TCP connections over high- LISTENING: In case of a server, waiting for a
bandwidth, high-latency networks. Well-tuned networks connection request from any remote client.
can perform up to 10 times faster in some
cases. However, blindly following instructions with
SYN-SENT: waiting for the remote peer to send back
understanding their real consequences can hurt the
a TCP segment with the SYN and ACK flags set.
performance as well.
('SYN-SENT' state is usually set by TCP clients)
TCP Bandwidth-delay product (BDP):
SYN-RECEIVED: waiting for the remote peer to
Bandwidth-delay product (BDP) is a term primarily
send back an acknowledgment after having sent back a
used in conjunction with the TCP to refer to the number
connection acknowledgment to the remote peer. ('SYN-
of bytes necessary to fill a TCP "path", i.e. it is equal to
RECEIVED' state is usually set by TCP servers)
the maximum number of simultaneous bits in transit
between the transmitter and the receiver.
ESTABLISHED: The port is ready to receive/send
High performance networks have very large BDPs. To data from/to the remote peer.
give a practical example, two nodes communicating
over a geostationary satellite link with a round trip FIN-WAIT-1: Indicated that the server is waiting for
delay of 0.5 seconds and a bandwidth of 10 Gbit/s can the application process on its end to signal that it is
have up to 0.5×1010 bits, i.e., 5 Gbit = 625 MB of ready to close.
unacknowledged data in flight. Despite having much
lower latencies than satellite links, even terrestrial fiber
links can have very high BDPs because their link FIN-WAIT-2: Indicates that the client is waiting for
capacity is so large. Operating systems and protocols the server's fin segment (which indicates the server's
designed as recently as a few years ago when networks application process is ready to close and the server is
were slower were tuned for BDPs of orders of ready to initiate its side of the connection termination)
magnitude smaller, with implications for limited
achievable performance. CLOSE-WAIT: The server receives notice from the
local application that it is done. The server sends its fin
TCP Buffers: to the client.
The original TCP configurations supported buffers of
up to 64 Kbytes (64 KB), which was adequate for slow LAST-ACK: Indicates that the server is in the
links or links with small round trip times (RTTs). process of sending its own fin segment (which indicates
Larger buffers are required by the high performance the server's application process is ready to close and the
options described below. server is ready to initiate its side of the connection
Buffering is used throughout high performance network termination).
systems to handle delays in the system. In general,
buffer size will need to be scaled proportional to the TIME-WAIT: Represents waiting for enough time to
amount of data "in flight" at any time. For very high pass to be sure the remote peer received the
performance applications that are not sensitive to acknowledgment of its connection termination request.
network delays, it is possible to interpose large end to According to RFC 793 a connection can stay in TIME-
end buffering delays by putting in intermediate data WAIT for a maximum of four minutes known as a MSL
storage points in an end to end system, and then to use (maximum segment lifetime).
automated and scheduled non-real-time data transfers to
get the data to their final endpoints [3] and [7]. CLOSED: Connection is closed.

P a g e | 10
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

TCP Connection establishment: considered a 2-way handshake since the FIN/ACK

To establish a connection, TCP uses a three- sequence is done in parallel for both directions.
way handshake. Before a client attempts to connect Some host TCP stacks may implement a half-duplex
with a server, the server must first bind to a port to open close sequence, as Linux or HP-UX do. If such a host
it up for connections: this is called a passive open. Once actively closes a connection but still has not read all the
the passive open is established, a client may initiate an incoming data the stack already received from the link,
active open. To establish a connection, the three-way this host sends a RST instead of a FIN. This allows a
(or 3-step) handshake occurs: TCP application to be sure the remote application has
read all the data the former sent—waiting the FIN from
1. SYN: The active open is performed by the the remote side, when it actively closes the connection.
client sending a SYN to the server. The client However, the remote TCP stack cannot distinguish
sets the segment's sequence number to a between a Connection Aborting RST and this Data Loss
random value A. RST. Both cause the remote stack to throw away all the
2. SYN-ACK: In response, the server replies data it received, but that the application still didn't read.
with a SYN-ACK. The acknowledgment Some application protocols may violate the OSI model
number is set to one more than the received layers, using the TCP open/close handshaking for the
sequence number (A + 1), and the sequence application protocol open/close handshaking; these may
number that the server chooses for the packet find the RST problem on active close.
is another random number, B.
3. ACK: Finally, the client sends an ACK back For a usual program flow like above, a TCP/IP stack
to the server. The sequence number is set to like that described above does not guarantee that all the
the received acknowledgement value i.e. A + data arrives to the other application.
1, and the acknowledgement number is set to TCP Resource usage:
one more than the received sequence number
i.e. B + 1. Most implementations allocate an entry in a table that
maps a session to a running operating system process.
At this point, both the client and server have received Because TCP packets do not include a session
an acknowledgment of the connection [6-8]. identifier, both endpoints identify the session using the
TCP Connection termination: client's address and port. Whenever a packet is received,
the TCP implementation must perform a lookup on this
The connection termination phase uses, at most, a four- table to find the destination process.
way handshake, with each side of the connection
terminating independently. When an endpoint wishes to The number of sessions in the server side is limited only
stop its half of the connection, it transmits a FIN packet, by memory and can grow as new connections arrive,
which the other end acknowledges with an ACK. but the client must allocate a random port before
Therefore, a typical tear-down requires a pair of FIN sending the first SYN to the server. This port remains
and ACK segments from each TCP endpoint. After both allocated during the whole conversation, and effectively
FIN/ACK exchanges are concluded, the terminating limits the number of outgoing connections from each of
side waits for a timeout before finally closing the the client's IP addresses. If an application fails to
connection, during which time the local port is properly close unrequited connections, a client can run
unavailable for new connections; this prevents out of resources and become unable to establish new
confusion due to delayed packets being delivered TCP connections, even from other applications.
during subsequent connections. Both endpoints must also allocate space for
A connection can be "half-open", in which case one unacknowledged packets and received (but unread)
side has terminated its end, but the other has not. The data.
side that has terminated can no longer send any data TCP Data transfer
into the connection, but the other side can. The
terminating side should continue reading the data until There are a few key features that set TCP apart
the other side terminates as well. from User Datagram Protocol:
It is also possible to terminate the connection by a 3-
Ordered data transfer — the destination host
way handshake, when host A sends a FIN and host B
rearranges according to sequence number.
replies with a FIN & ACK (merely combines 2 steps
into one) and host A replies with an ACK. This is
perhaps the most common method. Retransmission of lost packets — any cumulative
stream not acknowledged is retransmitted.
It is possible for both hosts to send FINs simultaneously
then both just have to ACK. This could possibly be Error-free data transfer.

P a g e | 11
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

Flow control — limits the rate a sender transfers data throughput during initial stage of the communication
to guarantee reliable delivery. The receiver continually (less than 50 Seconds). They conclude TCP Vegas will
hints the sender on how much data can be received be the best algorithm from the list.
(controlled by the sliding window). When the receiving In this work, they preferred six algorithms for
host's buffer fills, the next acknowledgment contains a 0 evaluation, because they are default algorithms in
in the window size, to stop transfer and allow the data several standard operating systems.
in the buffer to be processed. This work could extend in select few algorithms from
each of these four categories and will evaluate their
Congestion Control performance in MANET scenario. There are other
TCP Reliable transmission varying network parameters and metrics that the authors
are working on the same. Based on the results, it needs
TCP uses a sequence number to identify each byte of to be extending the further enhancement towards
data. The sequence number identifies the order of the specific application on MANETs [1].
bytes sent from each computer so that the data can be M.Jehan et. al. analyzed the TCP BIC and TCP Vegas
reconstructed in order, regardless of any fragmentation, congestion control algorithms and the performance of
disordering, or packet loss that may occur during these algorithms through NS2 simulator with proper
transmission. For every payload byte transmitted, the parameters. The tremendous growth of wireless
sequence number must be incremented. In the first two networks demands the need to meet different
steps of the 3-way handshake, both computers exchange congestion control algorithms for wireless network.
an initial sequence number (ISN). This number can be Usual transmission control protocol reduces its
arbitrary, and should in fact be unpredictable to defend performance by misinterpreting mobility losses due to
against TCP Sequence Prediction Attacks. TCP node motion as congestion losses in wireless mobility
primarily uses a cumulative acknowledgment scheme, network. The term network load is used to describe how
where the receiver sends an acknowledgment signifying much data will be transmitted over a connection with in
that the receiver has received all data preceding the the time duration. It is a gross measurement, taking the
acknowledged sequence number. The sender sets the total amount of data transferred in a given period of
sequence number field to the sequence number of the time as a rate, without taking into consideration the
first payload byte in the segment's data field, and the quality of the signal itself. The ratio of the number of
receiver sends an acknowledgment specifying the packets transmitted and the quantity of packets
sequence number of the next byte they expect to receive delivered to destination node within the time period.
[1] and [3] and [5]. They successfully evaluated the TCP BIC and TCP
Vegas congestion control algorithms using NS2
Performances Analysis of TCP Techniques
simulation tool. The performance of these two
M. Jehan, Dr. G.Radhamani & T.Kalakumari et. al. algorithms analyzed in ideal condition without any
analyzed six TCP Congestion Control Algorithms and cross traffic and any other additional flows. In this
their performance on Mobile Ad-hoc Networks small MANET scenario, the algorithm BIC provided
(MANET). More specifically, we describe the good throughput after 75 seconds but algorithm Vegas
performance behavior of BIC, Cubic, TCP Compound, provided stable and excellent result almost all over on
Vegas, Reno and Westwood congestion control the whole run time.
algorithms. The evaluation is simulated through This work is not sufficient reliable on performance on
Network Simulator (NS2) and the performance of these TCP congestion control. It could be improving the work
congestion control algorithms is analyzed with suitable evaluation with based on the as slow start and
metrics. The proposed simulations has been Congestion Avoidance [2].
successfully implemented and evaluated using NS-2 Marco Di Felice et. al. proposed a new paradigm to
simulator on a computer with Intel Core 2 Duo CPU provide additional spectrum utilization opportunities in
(T6400 processor @ 2.00 GHz) 2 GB of RAM. A wireless ad hoc networks. Recent research in this field
random wireless mobile ad hoc network topology was has mainly focused on devising spectrum sensing and
used for these experiments. They appraised the sharing algorithms, to allow an opportunistic usage of
performance of these congestion control algorithms in licensed portions of the spectrum by Cognitive Radio
very ideal condition without any cross traffic and any Users. However, it is also important to consider the
additional flows. In this small MANET scenario, the impact of such schemes on the higher layers of the
algorithm BIC (Binary Increase Congestion Control) protocol stack, in order to provide efficient end-to-end
provided good throughput after 75 seconds but data delivery. Since TCP is the de facto transport
algorithm Vegas provided stable and excellent protocol standard on Internet, it is crucial to estimate its
throughput almost all over on the whole run time. So it ability in providing stable end-to-end communication
move towards to the wrapping up that the algorithm over Cognitive Radio Ad Hoc Networks. The
Vegas will be the suitable algorithm for small and contributions of this paper are twofold. First, they
dynamic mobile ad hoc network scenario. Except proposed an extension of the NS-2 simulator to support
Vegas, all other assessed algorithms provided very poor realistic simulation of CRAHNs. The extension allows

P a g e | 12
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

modeling the activities of Primary Users (PUs), and the network protocols at different layers of the protocol
opportunistic spectrum management by CRUs in the stack.
licensed band. Second, it provided an accurate Simulation results show that sensing time interval and
simulation analysis of the TCP performance over type of PU activity play a critical role in deciding the
CRAHNs, by considering the impact of three factors: (i) TCP performance.
spectrum sensing cycle, (ii) interference from PUs and The simulation results highlight that transport protocols
(iii) channel heterogeneity. proposed for traditional wireless ad hoc networks might
They investigated the performance of different TCP not work well communication over Cognitive Radio Ad
variants over CRAHNs, i.e. TCP Reno, TCP New Hoc Networks. Since transport layer is still an explored
Reno, TCP with Selective Acknowledgment (TCP research area for Cognitive Radio Ad Hoc Networks
SACK) and TCP Vegas. The CRAHN environment is (CRAHNs), studying and understanding problems of
constructed. They assume that 5 PU (Primary Users) classical TCPs in the cognitive environment is
spectrum bands are present (i.e. B=5). Moreover, we fundamental to design novel transport protocol
assume that the number of CRU channels is equal to the solutions for Cognitive Radio Ad-hoc Networks [3].
number of primary bands (i.e. n=1), i.e. there is Jian Liu et. al. presented an approach where they
complete overlapping between CRU channels and PU implemented a thin layer between Internet protocol and
spectrum bands. This assumption makes easier to standard TCP that corrects these problems and
investigate the impact of PUs interference on CRUs maintains high end-to-end TCP throughput. The
using the same spectrum band. Individual spectrum protocol in Free BSD, and in this paper, we present
bands are occupied randomly and independently of each results from extensive experimentation done in an ad
other by PUs, according to the ON/OFF model. hoc network. The solution improves TCPs throughput
In the first case, the network is composed of 2 CRUs. A by a factor of 2–3. The goal in designing ATCP (Ad-
TCP/FTP connection is established between the static hoc TCP) was to provide a complete solution to the
CRUs. No mobility effect is considered at this stage. problem of running TCP over multihop wireless
They studied the performance of TCP under different networks. Specifically, they wanted to design a protocol
CRAHNs characteristics, e.g. the sensing time interval that has the following characteristics.
of CRUs, interference caused by PUs activity and 1) Improve TCP Performance for Connections set up in
bandwidth variation in an heterogeneous channel ad-hoc Wireless Networks. TCP performance is
environment. The choice of the single-hop scenario can affected by the problems of high BER and
be motivated firstly, the single-hop scenario is simple disconnections due to route re-computation or partition.
enough to understand the impact of CRAHNs In each of these cases, the TCP sender mistakenly
characteristics on the dynamics inside the TCP, while invokes congestion control. The appropriate behavior in
this might be difficult to investigate in multi-hop these cases ought to be the following.
topologies. Second, the single-hop topology constitutes  High BER: Simply retransmit lost packets without
a ―base case‖, from the point of view of protocol shrinking the congestion window.
performance. If we discover that a single parameter, e.g.  Delays due to Route Re-computation: Sender should
the sensing time interval, has a strong impact on TCP stop transmitting and resume when a new route has
performance, then this effect would be emphasized in a been found.
multi-hop environment by the presence of multiple Transient Partition: As above, the sender should stop
intermediate nodes between the source and the transmitting (because they did not want to flood the
destination pair. Moreover, although very simple, the network with packets that cannot be delivered anyway)
single-hop topology constitutes a realistic model for the until it is reconnected to the receiver.
evaluation of infrastructure-based CR networks, where Multipath Routing: In this case, when TCP at the
the mobile CRUs are attached to a fixed Cognitive Base sender receives duplicate ACKs, it should not invoke
Station (CBU). In the second scenario, they considered congestion control because multipath routing shuffles
multihop CRAHNs topologies, composed of 25 mobile the order in which packets are received.
CRUs. It varied the number of active TCP/FTP 2) Maintain TCP’s Congestion Control Behavior. This
connections. The multi-hop scenario is used to evaluate is an important goal because if losses are caused due to
end-to-end TCP performance when all the CRAHNs network congestion, they did not want the TCP sender
characteristics are considered. to assume that these losses were due to high BER and
In this paper, they addressed performance evaluation of continue transmitting. In this case, they want TCP to
TCP over Cognitive Radio Ad Hoc Networks shrink its congestion window in response to losses and
(CRAHNs). To this aim, we presented an extension of invoke slow start.
the NS-2 tool for the modeling and simulation of 3) Appropriate CWND Behavior. When there is a
CRAHNs. Their extension provides accurate modeling change in the route (e.g., a reconnection after a brief
of PUs activities and of CRUs spectrum management partition), the congestion window should be
functionalities, including spectrum sensing, decision recomputed.
and mobility schemes. Moreover, it provides facilities
to support cross-layer information sharing among

P a g e | 13
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

4) Maintain End-to-End TCP Semantics. It believe that a detrimental impact on TCP performance. To gauge
it is critical to maintain end-to-end TCP semantics in the impact of route changes on TCP performance, they
order to ensure that applications do not crash. derived an upper bound on TCP throughput, called the
5) Be Compatible with Standard TCP. This is necessary expected throughput. The TCP throughput measure
because we cannot assume that all machines deployed obtained by simulation is then compared with the
in an ad hoc network have ATCP installed. Thus, expected throughput.
machines with or without ATCP should be able to set They obtained the expected throughput as follows. We
up normal TCP connections with machines that may or first simulated a static (fixed) network of n nodes that
may not have ATCP. formed a linear chain containing n − 1 wireless hops.
Furthermore, applications running at machines with The nodes used the 802.11 MAC protocol for medium
ATCP Should not are aware of ATCP’s presence. access. Then, a one-way TCP data transfer was
Sometimes, it is likely that an ad hoc network may be performed between the two nodes at the ends of the
connected to wire line networks through access points. linear chain, and the TCP throughput was measured
In such situations, the sender or receiver of a TCP between these nodes. This set of TCP throughput
connection may lie in the wire line network with the measurements is analogous to that performed using
other end-point in the ad hoc network. It is important to similar (but not identical) MAC protocols.
ensure that TCP connections work normally in these This paper, investigated the effects of mobility on TCP
cases as well. The approach to the problem of performance in mobile ad hoc networks. Through
improving TCP’s performance while maintaining simulation, noted that TCP throughput drops
compatibility is to introduce a thin layer between TCP significantly when node movement causes link failures,
and IP called ATCP. In this paper, they presented a due to TCP’s inability to recognize the difference
solution to the problem of running TCP in ad hoc between link failure and congestion. Then they made
wireless networks. The solution is to implement a thin this point clearer by presenting several specific
layer between IP and ad-hoc TCP (called ATCP) that examples, one of which resulted in zero throughput, the
ensures correct TCP behavior while maintaining high other, in an unexpected rise in throughput with an
throughput. increase
This is done by putting TCP into persist mode when the in speed. We also introduced a new metric, expected
network is disconnected or when there are losses due to throughput, which provides a more accurate means of
high bit error. performance comparison by accounting for the
The Ad-hoc TCP’s performance is need to improve as differences in throughput when the number of hops
measured by the time to transfer large files. And also varies. We then used this metric to show how the use of
need to maintain end to end delay in ATCP congestion explicit link failure notification (ELFN) can
control behavior when there is network congestion [4]. significantly improve TCP performance, and gave a
Gavin Holland et. al. investigated the effects that link performance comparison of a variety of potential ELFN
breakage due to mobility has on TCP performance. protocols. In the process, they discovered some
Through simulation, they showed that TCP throughput surprising effects that route caching can have on TCP
drops significantly when nodes move, due to TCP’s performance.
inability to recognize the difference between link failure This work need intend to investigate ELFN protocols in
and congestion. They also analyzed specific examples, more detail, as well as the effects that other mobile ad
such as a situation where throughput is zero for a hoc routing protocols have on TCP performance. Also
particular connection. It introduce a new metric, more research is needed to better understand the
expected throughput, for the comparison of throughput complex interactions between TCP and lower layer
in multi-hop networks, and then use this metric to show protocols when used over mobile ad hoc networks, and
how the use of explicit link failure notification (ELFN) to find solutions to the problems caused by these
techniques can significantly improve TCP performance. interactions [5].
In this performance analysis, they set up a single TCP-
Reno connection between a chosen pair of sender and Martin Kohlwes et. al. reported the results from a long
receiver nodes and measured the throughput over the series of various measurements on the behavior of TCP
lifetime of the connection. over a UMTS wireless channel performed in two
They use throughput as the performance metric in this different UTMS networks. They conclude that at least
paper. The TCP throughput is usually less than in good conditions TCP throughput is close to
―optimal‖ due to the TCP sender’s inability to theoretical maximum, and that RTT is fairly stable.
accurately determine the cause of a packet loss. The Practically no packet losses were detected, and spurious
TCP sender assumes that all packet losses are caused by retransmissions were extremely rare. No performance
congestion. Thus, when a link on a TCP route breaks, benefit was observed when TCP retransmission timer
the TCP sender reacts as if congestion was the cause, modifications, such as the Eifel and FRTO algorithms
reducing its congestion window and, in the instance of a were used.
timeout, backing-off its retransmission timeout (RTO). Two types of TCP connections were used in the
Therefore, route changes due to host mobility can have measurements. First, connections between two UEs

P a g e | 14
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

were used. While this is an interesting case to study, It is unclear what takes place when networks become
most of the TCP traffic will probably be related to truly crowded [6].
accessing services in the fixed network. Thus, the bulk
of the measurements were performed in a scenario, Conclusion and Future Works
where a dedicated Linux-based server, running multiple
TCP-based services was contacted from a client running The Transmission Control Protocol (TCP) was designed
on a laptop connected to the UE. To study the behavior to provide reliable end-to-end delivery of data over
of the TCP state machine in more detail, small unreliable networks. In practice, most TCP deployments
modifications to the Linux kernel were done, enabling have been carefully designed in the context of wired
the logging of various internal variables, such as the networks. Ignoring the properties of wireless ad-hoc
congestion window size, the threshold value, and so on. Networks can lead to TCP implementations with poor
In addition to this data, TCP dump was used to obtain performance. In order to adapt TCP to the ad hoc
packet traces for further analysis during all environment, improvements have been proposed in the
measurements. literature to help TCP to differentiate between the
The control mechanism measurements were carried out different types of losses. Indeed, in mobile or static ad
UMTS offered a stable data channel for TCP to operate hoc networks losses are not always due to network
in. High packet loss rates and highly varying RTT congestion, as it is the case in wired networks or
commonly attributed to wireless channels were not wireless networks. After survey the techniques on TCP
observed. Thus, UMTS certainly has the potential to we conclude that the causes of TCP performance
offer high-speed internet connectivity using standard degradation in MANETs are due to many major
protocols and settings. Whether this potential continues problems. These problems are TCP is unable to
to be realized in commercial networks in the future is an distinguish between losses due to route failures and
interesting question. In the measurements, no other network congestion; TCP suffers from frequent route
users were to our knowledge present in the network. failures, End to End delay, and throughput. During the
Due to the design of the UMTS system, these survey, we also find some points that can be further
conclusions should also hold for networks used by not a explored in the future using advanced technique in
single, but a small group of users. The commercial feature extraction method and will improve the
realities, on the other hand, make it unlikely that these performance of TCP technique to achieve more
expensive networks run with only a few users per cell. efficient accuracy in network congestion, throughput
and reduce the end to end delay time.

After surveying different techniques we define the Advantages and Disadvantages of techniques in the table:

Techniques Advantages/ Merits Disadvantages /Future Improvement

Direction
The performance of these congestion This work could extend in select few
control algorithms in very ideal condition algorithms from each of these four categories
TCP Congestion Control without any cross traffic and any additional and will evaluate their performance in MANET
Algorithms, BIC, Cubic, flows. scenario. There are other varying network
Compound, Vegas, Reno, parameters and metrics that the authors are
Westwood. working on the same. Based on the results, it
needs to be extending the further enhancement
towards specific application on MANETs [1].

The performance of these algorithms This work is not sufficient reliable on

analyzed in ideal condition without any performance on TCP congestion control. It
cross traffic and any other additional flows. could be improving the work evaluation with
TCP Congestion Control In this small MANET scenario, the based on the as slow start and Congestion
Algorithms, MANET, BIC, algorithm BIC provided good throughput Avoidance [2].
Vegas after 75 seconds but algorithm Vegas
provided stable and excellent result almost
all over on the whole run time.

It provides facilities to support cross-layer The simulation results highlight that transport
information sharing among network protocols proposed for traditional wireless ad
protocols at different layers of the protocol hoc networks might not work well
Spectrum Management, stack. communication over Cognitive Radio Ad Hoc
Transmission Control Protocol Simulation results show that sensing time Networks. Since transport layer is still an
interval and type of PU activity play a explored research area for Cognitive Radio Ad
critical role in deciding the TCP Hoc Networks (CRAHNs), studying and

P a g e | 15
www.ijcait.com International Journal of Computer Applications & Information Technology
Vol. II, Issue I, January 2013 (ISSN: 2278-7720)

performance. understanding problems of classical TCPs in

the cognitive environment is fundamental to
design novel transport protocol solutions for
Cognitive Radio Ad-hoc Networks [3].
It presented a solution to the problem of The Ad-hoc TCP’s performance is need to
running TCP in ad hoc wireless networks. improve as measured by the time to transfer
The solution is to implement a thin layer large files. And also need to maintain end to
Ad-hoc TCP algorithm between IP and ad-hoc TCP (called ATCP) end delay in ATCP congestion control behavior
that ensures correct TCP behavior while when there is network congestion [4].
maintaining high throughput.

The use of explicit link failure notification This work need intend to investigate ELFN
(ELFN) can significantly improve TCP protocols in more detail, as well as the effects
performance, and gave a performance that other mobile ad hoc routing protocols have
comparison of a variety of potential ELFN on TCP performance. Also more research is
Explicit feedback algorithm, protocols. In the process, they discovered needed to better understand the complex
TCP Reno some surprising effects that route caching interactions between TCP and lower layer
can have on TCP performance. protocols when used over mobile ad hoc
networks, and to find solutions to the problems
caused by these interactions [5].

To study the behavior of the TCP state It is unclear what takes place when networks
machine in more detail, small modifications become truly crowded [6].
to the Linux kernel were done, enabling the
logging of various internal variables, such
as the congestion window size, the
TCP, FRTO algorithm threshold value, and so on. TCP throughput
is close to theoretical maximum, and that
RTT is fairly stable. Practically no packet
losses were detected, and spurious
retransmissions were extremely rare.

References

[1] M.Jehan, Dr.G.Radhamani, and T.Kalakumari,

―VEGAS: Better Performance than Other TCP [5] GAVIN HOLLAND, and NITIN VAIDYA,
Congestion Control Algorithms on MANETs‖, ―Analysis of TCP Performance over Mobile Ad Hoc
International Journal of Computer Networks (IJCN), Networks‖, Wireless Networks 8, 275–288, 2002,
Volume (3): Issue (2): 2011, pp. 151-158. Kluwer Academic Publishers. Manufactured in
Netherlands, pp. 275-278.
[2] M.Jehan, G.Radhamani, and T.Kalakumari,
―Experimental Evaluation of TCP BIC and Vegas in [6] Martin Kohlwes and Janne Riihiarvi and Petri
MANETs‖, International Journal of Computer Mahonen, ―Measurements of TCP Performance over
Applications (0975 – 8887) Volume 16– No.1, UMTS Networks in Near-Ideal Conditions‖, IEEE
February 2011, pp. 34-38. 2005.

[3] Marco Di Felice, Kaushik Roy Chowdhury, and [7] K. Chandran, S. Raghunathan, S. Venkatesan, and
Luciano Bononi, ―Modeling and Performance R. Prakash, ―A feedback based scheme for improving
Evaluation of Transmission Control Protocol over TCP performance in ad hoc wireless networks,‖ in Proc.
Cognitive Radio Ad Hoc Networks‖, MSWiM’09, 18th Int. Conf. Distributed Computing Systems,
October 26–29, 2009, Tenerife, Canary Islands, Amsterdam, The Netherlands, May 26–29, 1998, pp.
Spain.Copyright 2009 ACM 978-1-60558-616-8/09/10, 474–479.
pp. 4-9.
[8] S. Floyd, ―TCP and explicit congestion
[4] Jian Liu, and Suresh Singh,‖ ATCP: TCP for notification,‖ ACM Compute Communication. Rev.,
Mobile Ad Hoc Networks‖, IEEE JOURNAL ON vol. 24, no. 5, Oct. 1994, pp. 10-23
SELECTED AREAS IN COMMUNICATIONS, VOL.
19, NO. 7, JULY 2001, 1300-1315.

P a g e | 16