A novel algorithm for tcp timeout

International Journal of Computer Networks & Communications (IJCNC) Vol.7, No.3, May 2015
DOI : 10.5121/ijcnc.2015.7311 145
A NOVEL ALGORITHM FOR TCP TIMEOUT
CALCULATION OVER WIRELESS AD HOC
NETWORKS
Tareq T. Krishan1
, Adnan M. Al-Smadi2
1
Department of CIS, The University of Jordan-Aqaba, Jordan
2
Prof. Adnan M. Al-Smadi, Yarmouk University, Irbid, Jordan
ABSTRACT
To accommodate real-time multimedia applications that achieve Quality of Service (QOS) requirements in
a wireless ad-hoc network, a QOS control mechanisms is needed. Correspondence over such networks
must consider other aspects in regard to network properties; that the time it takes to send a message and
reach its destination faces different variables from one message to the other in a short time. Therefore, the
equation of calculating the time required to resend the message must be able to contain the worst case and
acknowledges different features for the network. The objective of this paper is to improve retransmission
time calculation adaptability when occurring using Transmission Control Protocol (TCP) over wireless Ad
hoc networks. Hence, it aims to obtain more accurate time to ensure retransmission time occurring in
accordance to the network environment variables efficiently.
KEYWORDS
TCP Computation Enhancement, Wireless networks, wireless Ad-hoc networks, Mobility, QOS for routing,
smart routing.
1.INTRODUCTION
Wireless ad-hoc networks do not require any fixed equipment or infrastructures such as routers,
switches, access points, base stations, or cables. Nodes are connected with each other through
radio signals to organize a network and transmit data from one node to another. Therefore
wireless ad hoc networks are considered as promising technology to establish a means of
communication where installation of network equipment and cables is not allowed due to some
difficulties, For instance high cost. For example, historic landmark, festival site, or when
conventional communication infrastructures are destroyed due to catastrophic disasters like
earthquake. In such situations, wireless ad-hoc networks are expected to accommodate real-time
multimedia flow for remote Monitoring, video conferencing, and Voice over IP (VoIP)
communications. Congestion may occur at one node or which generally affects QOS by routing.
Several papers on QOS routing have been proposed for wireless ad-hoc networks. QOS - Ad hoc
on-demand Distance Vectors (AODV) [4] is a per node available bandwidth estimation protocol
based on AODV. It estimates the available bandwidth using the number of packets exchanged
between nodes. On the other hand, Several Studies on retransmission adaptability have been
proposed to wireless ad-hoc networks and TCP timeout adaptively over wireless mobile ad hoc
networks [2] [5] to enhance retransmission’s timeout and reduce congestion which affects on the

146
overall throughput of the network. To our knowledge, there has been no work addressing the
problem of controlling time retransmission time in order to achieve performance gain in the
context of real-time multimedia applications over mobile wireless
ad hoc networks.
This paper proposes a new approach which ensures occurrence of retransmission in more adaptive
scenarios in case of congestion in one node or more, consistent with application QOS
requirements in a wireless ad hoc network. This paper focuses on new algorithm that has been
built and tested in the NS-2 simulation environment, where sequence has been applied in number
of nodes and movement speed through multiple test cases to ensure the stability of the low
percentage of delays with the increasing number of nodes. Thus, we are talking about the stability
of performance to coincide with unstable network.
The remainder of this paper is organized as follow. Problem statement is described in Section II.
Section III includes the literature review. Section IV presents a novel algorithm for TCP timeout
calculation over ad hoc networks. Simulation set up is described in section V. Proposed equations
are displayed in Section VI. Finally, conclusions and future work are included in Section VII.
2. PROBLEM STATEMENT
Recent years have seen rapid progress and evolution in the context of communication technology
and wireless networks. This was a result of the continuous development of accuracy in the
technology of Integrated Circuit (IC) industry. In addition, it provided a qualitative improvement,
which led to the acquisition of mobile devices with ultrafine sizes and lower costs have made
successes of a multifaceted in terms of the quality of performance in the services and applications
that provide user [1].
The evolution in wireless revolution governed by the quality of the applications and services
required, which is used by the studies and research related in wireless networks recently cares in
the quality of messaging and ideal transmission channels. The following are challenges faced by
wireless networks [2]:
• Limited capacity of transmission's frequencies Bandwidth (BW), where the wireless
network is characterized by lower capacity range of messaging and data rates offered
unlike the wired network which is characterized by a larger transmission capacity [3].
• long times to track Round Trip Time (RTT) , which tend wireless networks to own
delay, due to high relatively loss of packets more than wired networks . So, a renewed
tracks offering concept of mobility for the user to move free with enjoyment of the
services without interrupting the network [4].
• Energy consumption such as mobile computers, mobile devices and PDA, that managed
by cell power / battery. Therefore, there is need to reduce power consumption through
producing batteries less consumption [5], or smart routing algorithms.
The majority of the above factors impose themselves on applications and network services,
especially real world applications; for instance, video conferencing, that require to ensure a
sequential flow and order of packets to achieve the reliability and quality of the delivery to the
end users. Instant wireless environment is affected by several factors [6] including:

147
• Geographic locations. Variation in the flow of transmission over sight line of sight,
which is known as the Line of Sight (LOS) /Non Line of Sight (NLOS)
• High Latencies.
• Disconnect temporary connection.
• High degree of delay to deliver packages.
The low performance for TCP influenced by the properties of the wireless network that imposes
itself, which led to the research and investigation in recent years to make the TCP more adaptive
wireless work [7] [8].
Wireless networks are classified into two categories: namely: infrastructure networks and mobile
wireless networks. The most important characteristic of the second type from the rest of the other
networks is the amount of communications managed between nodes mobile in the absence of
standalone router, as it relates to nodes with each other. Mobile wireless network mechanism of
action is crystallized through temporary telecommunications and short-term with a plurality of
hops, each node acting as a guide to push the data packets and routed to hold other [9].
The basic idea of the decrease of the performance and quality of the TCP on the wireless network
is the inability to detect the position of the packet loss, which causes the loss attributable to a
variety of reasons not limited to network congestion, but for other things, including:
• Link Breakage disconnection, Handoff, Fading [10].
• Delay Ability (wireless mobile networks, for example, talking about the fullness of the
queue for a specified size and thus a drop of the packages).
• High error rate in the deciding one.
With respect to these causes for the loss of packets is generating reactions as a reduction of
inappropriate window transmission and reduce the period of the beginning of the transmission
and activation of control mechanisms in congestion reduces the productivity of the transport
protocol [11]. The causes of low productivity during the occurrence of transmission by using TCP
in a private mobile network illustrated as follows [7]:
• Effect of re- calculating the route to the source: by virtue of mobile nodes may change
tracks, and re-build new tracks that require being more obvious, that the process of
discovery of the new route takes time at length what makes TCP sender subject to a
waiting period longer, until the discovery of the new route.
• Effect of the division of the network: This problem occurs When stationed sender in part
of the network and recipient ensconce him into another part of the network. Packet cannot
reach to destination directly, sender still in a waiting period (Timeout) and by virtue of
the non-arrival notification access within the estimated round trip time (ERTT) of the
flight path and go back. Sender resends the package lost several times with doubling in
(timers); we talk about inflation in timeout.
• Effect of window congestion in mobile ad hoc networks (MANET): congestion window
indicate the percentage of accepted data for a particular party and by virtue of that, the
roads changed much capacitive given two ways will fluctuate as well.

148
Increasing or decreasing the number of mobile nodes over the path between the transmitter and
receiver with different speeds will renew the change in the values of the time of the flight path
default such as Sample Round Trip Time (SRTT); thus affected the values to guess the time of
the flight path (ERTT) in a manner sudden increases so led ultimately to possible inflation
processes and the value of the retransmission Timeout. Irregular variations for guess the round-
trip time and the possible delay caused by Retransmission Timeout (RTO) lowers the efficiency
of the TCP and hence the low quality of the service provided. That requires TCP to be more in
recognition of the properties of the network with the calculated values for retransmission.
Therefore, we talk about reducing effects by the changes to the values of renewable time in
virtual flight path. RTO lowers the efficiency of the TCP and hence the low quality of the service
provided.
Estimating times of renewable tracks the most important factor in the performance of the
transmission appears through the TCP, especially when dealing with change in the movement of
the node between the sender and the receiver. This will cause a sudden increase values to guess
the time of the flight path and thus increasing the value of timeout [12].
Temporary disconnection lead to wide difference on time values over virtual paths, in regards to
packets and approval its arrival. The big difference is known incremental sudden delay and thus a
“Spurious TCP Timeout” occurs, which reduce productivity and performance of the TCP [13].
This study aims to improve the calculation of time retransmission by the sender who uses the
TCP, which take the sender until the completion of the receiving instant, acknowledgment (ACK)
from the destination, and adapt to what can occur in this period under the conditions imposed by
the mobile network.
3. LITERATURE REVIEW
There are many previous studies regarding the adaptability of transmission protocol on wireless
networks in general. The studies focused specifically on operations to calculate values and guess
retransmission waiting time and the extent of vulnerability to the vagaries of flight paths, with
indirect talk about the nature of the problem, which circulated the current study. The paper in [16]
Presented algorithm EIFEL which made the TCP more durable on the wireless network, through
the removal of uncertainty for the re-messaging and being able to reduce the Spurious Timeout &
Spurious Fast Retransmission. Fahmy, et. Al. [9] suggested improving the performance of the
TCP packet loss in wireless scenarios, where showed that the collapse or loss of packets is not
due to congestion. However, it does not intersect with the proposed study in this paper in terms of
the impact of renewable infrastructure for the wireless network in terms of the movement of
mobile nodes on the performance of the TCP. Another study presented in [13] investigated the
subject of TCP Timeout on the wireless network and studied add additional delay is added
regularly to increase the contrast time flight path. This study has shown that the additional delay
added to the time of the flight path will increase the productivity of the TCP and reduce Spurious
Timeout formed as a result of the high variability of the values of the time of the flight path
Default. The big challenge faced by this study is that when the value of the guess is too long
Timeout lead to Underutilizing Network and when the value is very small guessing there will be
inflation in the retransmission. Kesselman and mansour [17] proposed that the value of the
retransmission ideal time depends on the RTT & TCP windows size. It also touched upon the
completion of the study to the time that the retransmission increases the productivity of the TCP,
noting that the time of retransmission is affected by the difference quotient of the flight path and
go back higher the value of the TCP windows size increases the value of the time of

149
retransmission. The paper also talked about the sense of Premature RTO occurrence send back
fast and early to coincide with the lack of an actual loss for the package, any package that comes
to be the status of the delay notice arrives late; Mini occurrence of resending unjustified. Costa
[19] suggested so-called FREEZ-TCP To alert the sender about the loss of signal expected in
transmission range to avoid the loss of packets and increasing the rate of delay their arrival, sends
a notice to the sender to freeze timers Instant (timers) to receive further notice acknowledges the
return signal strength Instant. Then, TCP sender proceeds into transmission.
The most important studies that provided a solution through the presence of the equations is
provided by Jacobson study [14]. The equations proposed provided optimal solutions in a manner
calculated wait time on wired networks to its stability with relatively lower rates of loss of
packets as the loss of packets on the wired network leads to congestion on the network. Theses
equations show the following:
• When data segments transmission began sequentially by the sender depending on TCP,
records the start time of the actual transmission to send the segment. When receiving the
notice of arrival (ACK) of that segment, the time of its receipt is also recorded. The
difference between those times called (SRTT), time that represent for send the packet and
receive delivery notification; that is,
SRTT = current time where the segment ACK received – Previous time where the segment
was sent to distance
• TCP calculates the estimated value of the time of route between the weighted average of
the value of time and track previously estimated the value of the flight path according to
the new equations [15]
‫ܴܶܶܧ‬ = ܽ ∗ ERTT + ሺ1 − aሻ ∗ SRTT (1)
where, ‘a’ is a constant value ranging between [0.8 – 0.9],
ERTT is the estimated round trip time
ܶ݅݉݁‫ݐݑ݋‬ = 2 ∗ ‫ܴܶܶܧ‬ (2)
Timeout is calculated as a function of the value of ERTT multiplied by the number (2), which
represents the process of Backoff. This process is trying to keep the network stable in the event
of a sudden burden on the network leads to the probability of loss of firmness [15].
‫݁ܿ݊݁ݎ݂݂݁݅ܦ‬ = ܴܵܶܶ − ‫ܴܶܶܧ‬ (3)
‫ܴܶܶܧ‬ = ERTT + ሺߜ ∗ ‫݁ܿ݊݁ݎ݂݂݁݅ܦ‬ሻ (4)
‫݊݋݅ݐܽ݅ݒ݁ܦ‬ = Deviation + ߜ ሺ |Diff | − Deviation ሻ (5)
Where, | | represent the absolute value
ܶ݅݉݁‫ݐݑ݋‬ = ߤ ∗ ERTT + ∅ ∗ Deviation (6)
Where, Deviation be between SRTT and ERTT, φ = 4 ideal value of inflation, µ = 1,
δ = 1/8 factor value Influenced with movement.

150
Equation (6) illustrates that if the difference is large, the deviation coefficient is dominant at the
expense of timeout. On the other hand, if the difference is small, then timeout approaches ERTT.
However, Jacobson algorithm applied to the mobile wireless network suffers from several
challenges for not adopting wireless scenarios taking into account the work in terms of the high
rates of loss of packets and mobility node then is repeated time values estimated route (Estimated)
between two mobile nodes difficult process. In addition, a deviation value for the time of route
from time to time leads the issue of the movement of nodes at different speeds. Also, application
of Equations (3-6) on the mobile wireless environment showed that the Timeout values are
increasing sharply due to the movement of nodes at different speeds either be oncoming traffic
from each decade, or vice versa.
The study in [12] refers to the direct reasons that leads to low effectiveness of Jacobson's
equations in the wireless environment through two scenarios relate to the movement of the
decade:
• Timeout will increase if two nodes moved away from each other, comparing to delay
with receive duplicated notifications.
• If two nodes closer to each other, timeout is supposed to decrease, but actually it
increases comparing to delay with receive duplicated notifications.
More recent study in this area is the study in [18]. This study presented several proposals to
amend the equation calculating the Timeout in the equations developed by Jacobson [14] to
become more suitable for mobile wireless networks taking into account the mobility of the node
and speed. Pointing out that the values of the times of route varies from time to time largely
because of the volatility in the node and speeds different. The study presented several proposals to
amend the Timeout calculation in Equation (6). The results obtained from Jacobson equation
applied in [18]. The deferments are decreasing between SRTT and ERTT, when number of hops is
lower and nodes speed is slow. The deferments are increasing between SRTT and ERTT, when
number of hops increases and nodes speed is fast.
The original form to the Jacobson equation is: ܶ݅݉݁‫ݐݑ݋‬ = ߤ ∗ ERTT + ∅ ∗ Deviation
So, we can summarize the results of Jacobson equation within wireless environment depending
on increasing and decreasing in the number of hops:
• When nodes movements near each other quickly, the number of hops will decrease and
thus SRRT reach to lowest level. After that each of ERTT and the impact of deviations
also reduced. So, ERTT will be the major factor through timeout computation with no
difference between SRTT and ERTT paths.
• Increasing the speed of the nodes makes the number of nodes that form the path of
communication is changing rapidly, thus increasing potential node makes changes at the
time of the flight path the default happen faster.
• In Jacobson’s Equation, Deviation value hits in the factor (Φ = 4) to enlarge the values
Timeout toward super value and therefore there is a big difference between the values of
the SRTT times and RTO.
Halasa in [18] has begun with the following modification on equation (6), so its amendment
produced the following equation:
ܶ݅݉݁‫ݐݑ݋‬ = μ ∗ ERTT + ሺሺ∅ ∗ Deviationሻ/nሻ (7)

151
Results can be summarized as follows:
• Talked about the worst case (Φ ∗ deviation) in terms of exception hit that value with the
number of hops, just Sufficiency in dividing by the number of nodes.
• With increase of the speed of movement and increase the number of nodes, increasing
changes in the estimated times for sending what makes the expression Deviation
dominant element in calculating the period Timeout.
Kajioka, et. Al. [20], proposed a QOS-aware routing mechanism for real-time applications. By
embedding bandwidth information in control messages of OLSRv2, a source node establishes the
logical path with the maximum available bandwidth. It confirmed that proposal could achieve
almost the same packet delivery ratio, the end-to-end delay, and the delay jitter as QOLSR in
general topology. But this study did not address the impact of the movement of nodes on the
accounts of TCP protocol regarding the time of re-transmission for packets and adaptive
computations that relating to this dilemma. Lee et. al., [21], explains how the network coding
(NC) is useful in the wireless network and evaluated various aspects of the performance of NC at
wireless ad hoc network. It found the impact each of (characteristic of MAC protocol, additive
power consumption by NC protocol and network conditions inappropriate for NC) on
performance evaluation. Therefore, the practical gain of NC is lower than the theoretical gain.
Also, this investigation does not refer to the TCP adaptability in regard to RTO and the effects of
mobility on overall network throughput.
4. A NOVEL ALGORITHM FOR TCP TIMEOUT CALCULATION
OVER AD HOC NETWORKS
The Transmission Control Protocol is a messaging Transfer Protocol to enable the reliability for
guaranteed delivery of data packets to their goals according to its harmonized multi-tasking
structure, together towards the optimization of its performance. On the other hand it uses the
Transmission Control Protocol timers’ retransmission (Retransmission Timers) as one of the
mechanisms of the main supports reliability to deliver data. It is the process of retransmission
based on the calculated value of the retransmission time (Retransmission Timeout), which spent
by the sender to re-send the missing package if elapsed time is estimated to reach the notice
received the package sent by the sender to the node. This value is calculated as a function of the
average time of the virtual transmission path, which represents the elapsed time between sending
data segment of the sender and the receipt of notice of delivery section node dispatcher. When
directly sender used for Transmission Control Protocol sent to the first data packet after the
adoption of a session contact with the receiving party through a mechanism called "Three Way
Hand Shake", then the time of the virtual transmission path time is calculated from the estimated
transmission path. Its value at the first package posted equal to the virtual transmission time.
That is, ERTT = SRTT (But it will vary later while sending the following packages due to
different placement of the node any movement) and then the value of the retransmission time will
be calculated even if the elapsed time is estimated to receive delivery of the package sent notice.
Then it a period of retransmission (Timeout) starts, which were calculated in advance to re-send
the package, lost successively until the receipt of the notice delivered means that the sender shall
remain for a period of time until the retransmission received notice of the arrival of the package
lost [16].
Methodology of the current study focused on the need for accuracy in regard to RTO computation
through making the processes of accounts more adaptive. Then, if these values are calculated in
inaccurate approach, it may lead to re-send the wrong information for lost packages.

152
The method of calculating the time of retransmission by Jacobson Karl’s equations has proven to
give good results on the network stable by virtue of the low rate of loss of firmness. The main
reason for the loss of packets on the wired network is congestion on the network [13]. Therefore,
the present study was to consider the factor of congestion can be included in the calculation of the
equation of time retransmission used in the Transmission Control Protocol and study its
behaviour over wireless networks for the mobile as if they contain the retransmission period of
time the value of the congestion at the path of an actor is not stable between the sender and the
future of the network.
Current study processes can be crystallized into the following functional steps:
1. Apply the proposed algorithm using TCL language via NS2 simulator and ongoing to start
with last formulation obtained by [18]; namely Equation (7) in this paper. This includes
timeout equation that consists of a factor for the number of nodes n. It should be
mentioned that the number of nodes used in the current study is 10.
• Analysis the behavior of Equation (7) via 15 nodes. That is expanding the number of
nodes to measure the overall throughput of this equation by increasing the number of
nodes. Thus, justifications are made for the amendment to the equation to bring a new
vision to the equation.
• Formulate the equations through present study that included the inclusion of the change in
the value of the congestion coupled with the number of nodes as the first vision is
included for the calculation of the equation of time retransmission within the present
study (15) node.
• After that, an optimal formulation is applied, with nodes increasing from 20 – 30 nodes.
Therefore, an ideal situation is applied to maintain a better level of delay on the network.
2. Make adjustments to the equation presented in [18], which include the inclusion of the
change in the congestion value at the node (Cwnd) coupled with the change in the
number of nodes n according to random velocities within the equation of calculating the
value of retransmission time.
3. Applying the modified equations within the increasing number of nodes between 20-30
nodes to reach to the optimal case, with stable in the delay level.
5. SET UP AND CONFIGURE THE PROPERTIES OF THE MOBILE WIRELESS
NETWORK SIMULATION (PREPROCESSING)
The process of creating the network’s properties within the simulation is crystallized at
the following steps:
First step
Configure each channel of transmission across the network and any coverage pattern which is
multiple destinations deployment on the network.
• Create a standard of physical specifications which is the participatory use of participatory usage of
transmission medium by end users including network frequency with ( 2.4 - 5 GHz ) with the
transfer rate of data across a network's nodes of (2 Mbps ) and throughput of ( 0.9 MB ).
• Next would be, configure a data link layer on the network and the type of sensor for wireless
messaging in each node.

153
• Create communication interface wireless that consists of PHY protocol to connect between data
link layer & physical layer as a communication protocol between these two layers through
network. That's because of its basic and direct role in running entities a peer to peer among users
of mobile network.
• Define a variable for the number of nodes that needs to be applied in the simulation.
• Define the size of the queue for packets at each node (what has a relationship with the flow
management and scroll in the nodes).
• Prepare the protocol that manages following processes (flow, scroll and dropping) for packets and
it in each of the nodes.
This means that we are talking about Queuing management methodology used in network, which
consists of Drop-tail protocol whose methodology is to achieve priority to push and scroll packets
on (FIFO) queue principle within incoming packets from different nodes either to pass it to
another node or to drop it. Then, packets from the queue that exceeds the size of the queue for
each node are dropped.
After that, define variable for routing protocols for mobile wireless network. Then define
simulation time in the study.
Finally, define 64 kbps transmission rate, we talk about the bandwidth of frequencies
transmission through the network.
Second step
Consist of configuring the nodes by assigning all the properties of the network that had been
prepared in First step to the nodes, starting from linking the number of nodes defined in the First
step to a channel network.
Then, the nodes in the network are positioned by selecting the initial sites for nodes that are
defined in the First step before starting the implementation of the Network Simulation.
Finally, implementation of random movement of nodes is requested when simulation is applied at
different speeds of the nodes.
Third step
• Consist of configuring sink Agent & TCP Agent who manages the transmission control between
two selected nodes as receiver and transmitter.
• Following this, achieving the linking of TCP Agent with the node selected as a source of
transmission to manage and control messaging sent from source.
• As well as achieving the linking of the Sink protocol Agent with nodes selected as a destination for
sending to control and manages instant Messaging with the recipient.
• Then comes the actual linking between messaging management protocols for both source and
Destination.
• Following this, defining and configuring communication and interaction management for
intermediate nodes that represent path between specific source and specific target.
• And then, implementing the actual communication and interaction management by creating
Constant Bit rate Application that is linked with TCP Agent as the application that runs the
adjustment of traffic to queue the packets in the path from the source node, Through several
criteria like determining the size of the packet which is 2321 Byte also like determining the data
transfer speed rate through the path between source and target which is 600 Kbps.

154
The following is the preprocessing steps mentioned above, which represents the Characteristics to
the environment of an Ad hoc network during the study.
Figure 1. Preprocessing steps
1. Step One
Set channel-type Wireless Channel
Set Propagation –Model Two Ray ground
Set Mac-type IEEE 802.11
Set Link-Layer-type LL
Set Antenna-model Omni Antenna
Set Network-Interface-Type Phy/wireless
Set No. Of Mobile Nodes n
Set Max-packet-In-Ifq 10000
Set Interface-Queue-type Drop tail/Priority Queue
Set Routing-Protocol Var(rp),
where Var(rp) is a variable Selected by the
user. ( Aodv, Dsdv , Dsr ….)
Set Simulation Time 150 Second
Set Mac-Data-Rate 2 MB/s
Set Mac-Bandwidth-Rate 64 Kb/s
2. Step Two
Set Nodes-Configuration as defined in Step 1
Set Initial position For the Nodes, Defined
Previously
Start Mobile-Nodes Motion randomly
3. Step Three
Set Tcp Agent, for the sender
Set Tcp Sink, for the receiver

155
6. PROPOSED EQUATIONS
ܶ݅݉݁‫ݐݑ݋‬ = ‫ܴܶܶܧ‬ + ݊ ∗
஽௘௩௜௔௧௜௢௡
Φ∗஼௪௡ௗ
(8)
Where, ERTT is the estimated round trip time, Cwnd is changes the value of the congestion with
the change of the number of nodes obtained by TCP agent, n is number of nodes, Deviation is
the deviation between SRTT and ERTT, ΦΦΦΦ = ૝ is Ideal value of inflation, used by Jacobson/
Karels [22].
The application of equation's algorithm (8) as shown in the Figure 2, the following result has
been obtained from simulation as shown in Figure 3.
Algorithm
Procedure Plot ( )
Set SRTT = 0; initial value.
Set ERTT = 0; initial value.
Set Deviation = 0; initial value.
Set δ = 1/8
Set µ = 1
Set Ф = 4
Get SRTT from TCP agent
Set ERTT = SRTT
Set Time = 0.01 ms
LOOP
i: 0.01 : 150
Get SRTT from TCP agent.
Get Cwnd from TCP agent.
Set ‫݁ܿ݊݁ݎ݂݂݁݅ܦ‬ = ܴܵܶܶ − ‫ܴܶܶܧ‬
Set ‫ܴܶܶܧ‬ = ‫ܴܶܶܧ‬ + ሺ ࢾ ∗ ‫݁ܿ݊݁ݎ݂݂݁݅ܦ‬ሻ
Set ‫݊݋݅ݐܽ݅ݒ݁ܦ‬ = ‫݊݋݅ݐܽ݅ݒ݁ܦ‬ + ࢾ ∗ ሺ|‫|݁ܿ݊݁ݎ݂݂݁݅ܦ‬ − ‫)݊݋݅ݐܽ݅ݒ݁ܦ‬
Set ܶ݅݉݁‫ݐݑ݋‬ = ‫ܴܶܶܧ‬ + ݊ ∗
஽௘௩௜௔௧௜௡
ః ∗ ஼௪௡ௗ

Commit file F0 = RTT
Commit file F1 = Timeout
Set ܶ݅݉݁ = ܶ݅݉݁ + 0.01
End LOOP
Execute X Graph For Files F0 , F1
Execute
Figure 2. The algorithm for the first modification.

156
Figure 3. Applying first modification using AODV via 15 nodes.
ܶ݅݉݁‫ݐݑ݋‬ = ‫ܴܶܶܧ‬ + φ ∗
஽௘௩௜௔௧௜௢௡
௡∗஼௪௡ௗ
(9)
where, ERTT is the estimated round trip time, Cwnd is changes the value of the congestion with
the change of the number of nodes Obtained by TCP agent, n is number of nodes, Deviation is
The deviation between SRTT and ERTT, ΦΦΦΦ = ૝ is Ideal value of inflation, used by Jacobson/
Karels.
Applying equation (9) to the algorithm in Figure 2, with change in the equation, the following
result has been obtained using simulation as shown in Figure 4.
Figure 4. Applying second modification using AODV via 15 nodes.
ܶ݅݉݁‫ݐݑ݋‬ = ‫ܴܶܶܧ‬ +
஽௘௩௜௔௧௜௢௡
(௡∗஼௪௡ௗ)
(10)
Where, ERTT is Estimated round trip time, Cwnd is Changes the value of the congestion with
the change of the number of nodes Obtained by TCP agent , n is number of nodes, Deviation is
The deviation between SRTT and ERTT, ΦΦΦΦ = ૝ is ideal value of inflation, used by Jacobson/
Karels.
In the application of equation's algorithm (10) using Figure 2, with change in the equation, the
following result has been obtained using simulation as shown in Figure 5.

157
Figure 5. Applying third modification using AODV via 15 nodes.
Figure 6. Applying third modification using AODV via 20 nodes.
ܶ݅݉݁‫ݐݑ݋‬ = ‫ܴܶܶܧ‬ +
஽௘௩௜௔௧௜௢௡
ఒ
(11)
where, ERTT is Estimated round trip time, Deviation is The deviation between SRTT and
ERTT , ߣ = Ф ∗ ‫݀݊ݓܥ‬ ∗ ݊, Cwnd is Changes the value of the congestion with the change of the
number of nodes obtained by TCP agent, n is number of nodes, ΦΦΦΦ = ૝ is Ideal value of
inflation, used by Jacobson/ Karels.
In the application of equation's algorithm (11) using Figure 2, with change in the equation, the
following result has been obtained using simulation as shown in Figure 7.
Figure 7. Applying fourth modification using AODV via 20 nodes.

158
A comparison of the retransmission timeout (RTO) using the original equation in [18]
AND the modified proposed equation in this paper is shown in Table 1.
Table 1. Comparison between the results using original and proposed (modified) equations
Method Number of Nodes RTO in msec
Original equation in [18]
Equation (11)
10 38.5584
Modified equation
Equation (12)
15 26.3908
Modified equation
Equation (15)
30 26.6854
Hence, the retransmission timeout has maintained a relatively good level with doubling
the number of nodes from 15 to 30 nodes. That is, the difference in the rate of RTO
during incrementing the number of nodes is 0.2946 msec. Therefore, RTO is not affected
by increasing the number of nodes, thus preserving a low level of delay in the network.

159
7. CONCLUSIONS AND FUTURE VISIONS
The paper presented an algorithm to improve the retransmission time over ad hoc networks. The
algorithm is a modification to the formula developed by [18]. Several experiments were
performed taking different number of nodes. A comparison was made between the formula in
[18] and the proposed modification. The results obtained shows that the proposed method
outperformed the one in [18]. The retransmission time was about the same when we went from
10 to 30 nodes using the proposed algorithm.
We suggest that different studies should be considered to introduce new mechanisms that achieve
smart and more effective routing, with taking into account the mobility Characteristics.
ACKNOWLEDGMENTS
First Author would like to thank Prof. Adnan Al-Smadi for his guidance and continuous support
through his experience and knowledge. The authors would also like Recognize that the work has
relied on the objectivity and transparency in all aspects of this research.
REFERENCES
[1] Shondip, S., " Localized Demand Driven Routing for Mobile Ad hoc Wireless Networks " , Master
Thesis, Computer science Department, University of Bristol , UK, pp. 39-47 , Feb 2005 .
[2] Crow, B., Widjaja, I., Kim, J., and Sakai, P., " IEEE 802.11 Wireless Local Area networks " , IEEE
Commun.Mag , Vol.35 , No.9 , pp. 116-126 , September 1997.
[3] Jamieson, K., and Balakrishnan, H., " PPR : Partial Packet Recovery for Wireless Networks ". In
Proc. Of ACM SIGCOM, Kyoto , Japan , August 2007, pp. 409–420 .
[4] Xu, S., and Saddawi, T., " Does the IEEE 802.11 Mac protocol work well in multihope wireless ad
hoc Networks? " , IEEE communication Magazine , Vol.39 , No.6 , pp. 130-137 , 2001.
[5] Srinivas, A., and Modiano, E., " Minimum Energy Disjoint Path Routing in Wireless Ad Hoc
Networks " , ACM Mobicom , pp. 122-133 , 2003 .
[6] Balakrishnan, H., Seshan, S., Amir E., and Katz, R., " Improving TCP/IP Performance Over
Wireless Networks " , Proceedings of Mobicom , Berkeley , CA , United States , pp. 2-11 , 1995.
[7] Hanbali, E., Nain, p., " A survey of TCP Over Mobile Ad hoc Networks " , Theme COM , Vol.7 , pp.
22-36 , May 2004 .
[8] Moustafa, k., " TCP over Mobile Ad hoc Access Networks “, Master Thesis ,Carleton University
Ottawa, Ontario, December 2002.
[9] Fahmy, S., Probhakar, S., Avasarala, S., and Younis O., " TCP Over Wireless Links : Mechanisms
and implication " , accepted for publication , Computer Science , Purdue University , pp. 1-9 , 2002 .
[10] Duel-Hallen, A., Hu, S., and Hallen, H., " Long-Range Prediction of Fading Signals:Enabling
Adapting Transmission for Mobile Radio Channels ", IEEE Signal Processing , Vol.17 , no.3 , pp. 1-
4 , May 2000 .
[11] Natani, J., Mohsin, M., and Sharma, V., " TCP For wireless networks ", Computer science Program ,
University of Texas at Dallas , pp. 16-20 , 2001 .
[12] Kandah, F., " Performance Evaluation of TCP's Fast Retransmission over IEEE 802.11's DCF " ,
Master Thesis , Computer Science Department , University of Jordan , Amman , June 2005 .
[13] Klein, E., Leung, K., Parkinson, R., and Samuel, G., “Avoiding Spurious TCP Throughputs in
Wireless Networks by Delay Injection”, Global Telecommunications Conference. GLOBECOM ’04.
IEEE, Vol. 5, no.29, pp. 2754 – 2759, December 2004.
[14] Jacobson, V. “Congestion Avoidance and Control”, Proc. of the Symposium on Communications
Architectures and Protocols (SIGCOMM), Stanford, CA, PP. 314-329, Augest 1988.
[15] Peterson, L., and Davie, B. " Computer Networks : A System Approach " , Morgan Kauffmann,
Vol.3, no.4, pp. 101-115, 2003.

160
[16] Ludwig, R., and Sklower, K., “The Eifel Retransmission Timer ". Computer Communication Review,
Vol.3, no.3, pp. 17-27, 2000.
[17] Kesselman , Y. Mansour " Optimizing TCP Retransmission Timeout " P. Lorenz and P. Dini (Eds.):
ICN 2005, LNCS 3421, Springer-Verlag Berlin Heidelberg, pp. 133–140 , 2005.
[18] Halasa, R., " Studying TCP's Timeout Adaptivity Over Wireless Mobile Ad hoc Networks ", Master
Thesis, University of Jordan, Amman, August 2006.
[19] Da costa, G., " Freeze TCP with timestamps for fast packet loss recovery after disconnections ",
Science direct magazine , Private Bag 4800, Department of Electrical and Computer Engineering,
University of Canterbur , Christchurch, New Zealand, pp. 1792-1796, 2003 .
[20] Kajioka, S., Wakamiya, N., Satoh, H., Monden, K., Hayashi, M., Matsui, S., Murata, M., “ A
QoS-aware routing mechanism for multi-channel multi-interface ad-hoc networks ” . Elsevier – Ad
hoc networks, Vol. 9, no.5, pp. 911 – 927, July 2011.
[21] Lee, K-H., Cho, S., Kim, J-H., “ performance evaluation of network coding in IEEE 802.11
wireless ad hoc networks ” Elsevier – Ad hoc networks, Vol.16 , pp. 131 – 141 , May 2014.
[22] Krishan, T., “Timeout Adaptability for TCP Retransmission Ad Hoc Networks”, Master Thesis, Al
Al-Byte University, Jordan, May 2009.
AUTHOR’S
Tareq t. Krishan Received the BSc. Degree in Computer science from Jerash private
university, Jerash, Jordan in 2004, the MSc. Degree from Al- al Bayt University, al
mafraq, Jordan in 2009. Presently, He is a lecturer with the Computer information
systems department, The University of Jordan – Aqaba. His current research interests
include mobile ad hoc networks, TCP adaptability over wireless networks and mobility.
Adnan M. Al-Smadi received B.S. degree (Magna Cum Laude) and the M.S. degree in
electrical engineering from Tennessee State University, Nashville, TN, USA in 1987
and 1990, respectively. He received the Ph.D. degree in electrical and computer
engineering from Vanderbilt University, Nashville in 1995.
From 1989 to 1991, he was an instructor of Mathematics at Tennessee State University, and from 1991 to
1992, he was an instructor of Mathematics at Fisk University, Nashville. From 1992 to 1995, he was an
instructor of Aeronautical and Industrial Technology at Tennessee State University. In addition, from 1995
to 1997, he was an adjunct Assistant Professor of Electrical and Computer Engineering at Vanderbilt
University. From 1995 to 1997, he served as Interim Department Head of the Aeronautical and Industrial
Technology Department at Tennessee State University. From 1997 to 2006, he joined Hijjawi College for
Engineering Technology as a faculty member of Electronics Engineering at Yarmouk University, Jordan.
From 2002 to 2004, he served as the Chairman of the Department of Electronics Engineering at Yarmouk
University. From 2004 to 2006, Professor Al-Smadi served as the Dean of Hijjawi College for Engineering
Technology. From 2006 to 2010, he was on sabbatical leave as a professor of Computer Science and the
Dean of Prince-Hussein Bin Abdullah College for Information Technology at Al Al-Bayt University in
Jordan. From May 2012 to September 2012, he was the Head of of Quality Assurance & Accreditation
Department at Yarmouk University. From 2009 to 2014, Prof. Al-Smadi served as a Member of the
"Recognition of non-Jordanian Universities & Certificates' Equivalency," Committee at the Ministry of
Higher Education and Scientific Research, Jordan. Since 2013, he has been serving as a "Member of the
Board of Trustees," Jerash Private University, Jordan. In addition, Professor Al-Smadi is serving now as the
Dean of Hijjawi College for Engineering Technology.
Professor Al-Smadi is a Senior Member of the IEEE and a member of Eta Kappa Nu in the USA.

A novel algorithm for tcp timeout

Recommended

More Related Content

What's hot (20)

Viewers also liked (19)

Similar to A novel algorithm for tcp timeout (20)

More from IJCNCJournal (20)

Recently uploaded (20)

A novel algorithm for tcp timeout