Performance Comparison of TCP variants over

AODV, DSDV, DSR, OLSR in NS-2

Md. Monzur Morshed* 1, 2, Meftah Ur Rahman 2, 3, Md. Habibur Rahman 1, 2, Md. Rafiqul Islam1
{[email protected], [email protected], [email protected]},
[email protected],
[email protected],
[email protected]
Department of Computer Science
American International University-Bangladesh1
SCICON & TigerHATS Research Team-Bangladesh2
George Mason University-USA3

Abstract—The cardinal concept of TCP development was to carry

comparative performance of four TCP variants over four
data within the network where network congestion plays a vital
role to cause packet loss. On the other hand, there are several different routing protocols.
other reasons to lose packets in Mobile Ad Hoc Networks due to
fading, interfaces, multi-path routing, malicious node and black II. OVERVIEW OF SIMULATED ROUTING PROTOCOLS
hole. Along with throughput, fairness of TCP protocols is
important to establish a good communication. In this paper, an A. AODV
empirical study has been done by simulation and analysis of TCP
Ad hoc On-Demand Distance Vector Routing protocol
variations under AODV, DSDR, DSR, OLSR routing protocol. In
(AODV) is an on demand routing protocol. To find a route to
our simulation, we studied multiple variations of TCP, such as
Reno, New-Reno, Vegas, and Tahoe. The simulation work has the destination, the source node floods the network with
been done in NS2 environment. Based on the simulation results RouteRequest packets. The RouteRequest packets create
we carried out observations for different TCP packets under temporary route entries for the reverse path through every
several QoS metrics such as drop, throughput, delay, and jitter. node it passes in the network. When it reaches the destination
a RouteReply is sent back through the same path the
Keywords-MANET; AODV; DSDV; DSR; OLSR; Tahoe; RouteRequest was transmitted. Every node maintains a route
Reno; New-Reno; Vegas; NS2 table entry which updates the route expiry time [3, 4, 5].

I. INTRODUCTION B. DSDV
TCP is one of the most popular end-to-end protocols offers Destination-Sequenced Distance Vector routing protocol
reliable connection and compatible for both in wired and (DSDV) is one of the most well known table-driven routing
wireless networks. Unlike wired links, wireless radio channels algorithms for MANETs which is based on the Distributed
are affected by many factors that may lead to high levels of Bit Bellman-Ford algorithm [6, 7]. DSDV routing protocol
Error Rate (BER) [1]. Though, TCP does not have the maintains a routing table that lists all available destinations,
functionality to determine the packet loss where the reasons the number of hops to reach the destination and the sequence
can be network congestion, channel errors, link failure, number assigned by the destination node. The routing table
fading, interfaces, multi-path routing, malicious nodes and update can be sent in two ways: a "full dump" or an
black hole etc., it has been the dominant transport-layer incremental update. A full dump sends the full routing table to
protocol providing reliable byte stream delivery between end- the neighbors and could span many packets whereas in an
host applications with mechanism of connection management, incremental update only those entries from the routing table
congestion control, flow control, and error control [2]. are sent that have a metric change since the last update.

MANET or mobile Ad hoc network is a collection of C. DSR

mobile nodes that are dynamically and arbitrarily located
where interconnections between nodes are capable of Dynamic Source Routing (DSR) is a kind of source
changing on a continual basis. In order to facilitate routing based on two procedures: route discovery and route
communication within the network, a routing protocol is used maintenance [8]. Route Discovery is the mechanism by which
to discover routes between nodes. After path establishment, a node S wishing to send a packet to a destination node D
either connection oriented protocol TCP or connection less obtains a source route to D. Route Discovery is used only
protocol UDP is necessary to transfer the actual data packets. when S attempts to send a packet to D and does not already
Due to its reliability, TCP and its variants play a crucial role in know a route to D. Route Maintenance is the mechanism by
data transfer over MANET. Though similar studies have been which node S is able to detect, while using a source route to D,
carried out earlier but this paper provides a succinct view of if the network topology has changed such that it can no longer
the use its route to D because a link along the route no longer

This study source was downloaded by 100000824502131 from CourseHero.com on 05-05-2021 13:15:58 GMT -05:00

https://www.coursehero.com/file/52872281/Performance-comparison-of-TCP-variants-opdf/
 works. When Route Maintenance indicates a source route is (up to 20%) at high node mobility adopting Random way-
broken, S can attempt to use any other route it happens to point model. Wei, Yeung and Hai [16] implemented TCP
know to D, or can invoke Route Discovery again to find a new performance comparison in Random Waypoint (RW) and
route. Route Maintenance is used only when S is actually Social Network (SN) mobility models and showed that SN
sending packets to D [9]. outperforms RW. Yoon, Liu and Noble [17] claimed Random
Waypoint is considered harmful as their findings showed that
this model fails to provide a steady state in that the average
D. OLSR nodal speed consistently decreases over time, and therefore
Optimized Link State Routing protocol (OLSR) [10] is should not be used directly for simulation.
based on link state algorithm and it is proactive in nature.
OLSR is an optimization over a pure link state protocol as IV. OVERVIEW OF TCP VARIENTS
topological change cause the flooding of the information to all
available hosts in the network. In OLSR, each node uses the A. TCP Tahoe
most recent information to route a packet. Each node in the
network selects a set of nodes in its neighborhood, which Tahoe refers to the TCP congestion control algorithm
retransmits its packets. This set of selected neighbor nodes is which was suggested by Van Jacobson in his paper [18]. This
called the multipoint relays (MPR) of that node. These are implementation added a number of new algorithms and
used to reduce flooding of broadcasts. OLSR uses two kinds refinements to earlier implementations. The new algorithms
of the control messages: Hello and Topology Control (TC) include Slow-Start, Congestion Avoidance, and Fast
[10]. Retransmit.
TCP packet transmissions are clocked by the incoming
RELATED RESEARCH
III. acknowledgements. However there is a problem- when a
Ad hoc network routing protocols have been well-studied connection first starts up it needs to have acknowledgements
in the past few years. TCP variants such as NewReno, Reno, so we need to have data in the network and to put data in the
Tahoe, Vegas are often used to evaluate the performance of network we again need acknowledgements. To get around this
MANETs routing protocols. circularity, Tahoe suggests that whenever a TCP connection
starts or re-starts after a packet loss it should go through a
Tayade and Sharma [11] used TCP variants such as Tahoe, procedure called ‘slow-start’.
Reno, Lite to compare the performance of DSR where they
found that TCP Reno was the best among Tahoe, Reno and For congestion avoidance, Tahoe uses ‘Additive Increase
Lite, it can be taken into consideration based on node speeds Multiplicative Decrease’. A packet loss is taken as a sign of
versus throughput and packet loss.TCP Lite suitable for the congestion and Tahoe saves the half of the current window as
variation in no. of node for both throughput and packet loss. a threshold value. It then set the congestion window to one and
starts slow start until it reaches the threshold value. After that
In reference [12], an improved TCP New Reno (ABRA it increments linearly until it encounters a packet loss. Thus it
New Reno) was proposed for mobile ad-hoc networks using increases its window slowly as it approaches the bandwidth
calculations of New Retransmission Time out, to improve capacity.
performance in terms of congestion control and implemented
in a Mobile Ad-hoc Network under QualNet 4.0 simulator. B. TCP Reno
ABRA New Reno performs well under the varying conditions TCP Reno is the most widely adopted Internet TCP
of high density node, high node speed and pause time, because protocol. It retains the basic principle of Tahoe, such as slow
of proper utilization of time, optimal paths between nodes, starts and the coarse grain re-transmit timer. However it adds
optimal bandwidth exploitation and less packet delay. some intelligence over it so that lost packets are detected
Elaarag in [13] showed that TCP protocol performs earlier and the pipeline is not emptied every time a packet is
efficiently if a link layer protocol suitable for mobile networks lost. It employs four transmission phases: slow start,
is used. The use of link layer protocols to provide an congestion avoidance, fast retransmit, and fast recovery. When
acceptable error performance over the wireless connection is packet loss occurs in a congested link due to buffer overflow
now a standard industrial practice while link layer protocols in the intermediate routers, either the sender receives three
can efficiently provide reliability, transport layer protocols can duplicate acknowledgments or the sender’s retransmission
be designed to efficiently deal with handoff and timeout (RTO) timer expires. Thus, TCP Reno requires that
disconnections. we receive immediate acknowledgement whenever a segment
is received. If we receive a number of duplicate
Lee, Ahn and Campbell in [14] showed the INSIGNIA acknowledgements then that means that sufficient time have
system not only improves TCP goodput but also showed passed and even if the segment had taken a longer path, it
improved service quality over all mobility conditions. At high should have gotten to the receiver by now. There is a very high
mobility, TCP flows often decrease their window segment size probability that it was lost. So Reno suggests an algorithm
to the minimum due to packet losses resulting from lack of called ‘Fast Re- Transmit’ [19, 20].
connectivity or congestion experienced in the network. More
congestion points are observed under higher mobility.
Kim, Bae and Toh [15] used ns-2 and their proposed TCP-
Vegas-ad hoc showed a better performance than the standard
TCP-Vegas. In particular, it improved the performance greatly

This study source was downloaded by 100000824502131 from CourseHero.com on 05-05-2021 13:15:58 GMT -05:00
C. TCP New-Reno New-Reno is a slight modification over TCP-Reno. It is able
https://www.coursehero.com/file/52872281/Performance-comparison-of-TCP-variants-opdf/
 to detect multiple packet losses and thus is much more VI. SIMULATION DESCRIPTION
efficient that RENO in the event of multiple packet losses.
Like Reno, New-Reno also enters into fast-retransmit when it TABLE I. SIMULATION PARAMETERS
receives multiple duplicate packets; however it differs from
RENO in that it doesn’t exit fast-recovery until all the data Method Value
which was out standing at the time it entered fast-recovery is
acknowledged. Thus it overcomes the problem faced by Reno Channel type Channel/Wireless channel
of reducing the congestion window size multiples times. The Radio-propagation model Propagation/Two ray round
fast-transmit phase is the same as in Reno. The difference is
the fast-recovery phase which allows for multiple re- Network interface type Phy/wirelessphy
transmissions in new-Reno [21]. TCP New-Reno exits fast MAC type Mac/802.11
recovery after receiving acknowledgement of all
unacknowledged segments. It then sets congestion window Interface queue type Queue/Drop Tail
size to slow start threshold and continues the congestion
Link Layer Type LL
avoidance phase [22].
Antenna Antenna/omni antenna
D. TCP Vegas
Maximum packet in ifq 50
Vegas is a TCP implementation which is a modification of
Reno. It builds on the fact that proactive measures to Area (m×m) 1000×1000
encounter congestion are much more efficient than reactive
Number of mobile nodes 16
ones. It tries to get around the problem of coarse grain
timeouts by suggesting an algorithm which checks for TCP (NewReno, Reno,
timeouts at a very efficient schedule. Also it overcomes the Source type
Tahoe, Vegas)
problem of requiring enough duplicate acknowledgements to
detect a packet loss, and it also suggests a modified slow start Simulation Time 100 seconds
algorithm which prevents it from congesting the network. It Routing protocol AODV, DSDV, DSR, OLSR
detects congestion before the packet losses occur. However it
still retains the other mechanism of Reno and Tahoe, and a
packet loss can still be detected by the coarse grain timeout of VII. QOS METRICS AND SIMULATION RESULT
the other mechanisms fail. The three major changes induced
by Vegas are: New Re- Transmision Mechanism, Congestion We used different parameter of QoS metrics such as delay,
Avoidance by means of comparing sending and expected rate jitter, packet drop and throughput to understand the behavior of
and Modified Slow-start [23]. TCP for AODV, DSDV, DSR, and OLSR Routing Protocol under
the time period of 100 sec in both static and dynamic network
simulation scenario.
V. SIMULATION TOPOLOGY
There Simulation environment consists of 16 wireless A. Drop
mobile nodes which are placed uniformly and forming a The routers might fail to deliver (drop) some packets if
Mobile Ad-hoc Network, moving about over a 1000 × 1000 they arrive when their buffers are already full. Some, none, or
meters area for 40 seconds of simulated time. We have used all of the packets might be dropped, depending on the state of
standard two-ray ground propagation model, the IEEE 802.11 the network, and it is impossible to determine what will
MAC, and Omni-directional antenna model of NS2. We have happen in advance. The receiving application may ask for this
used AODV, DSDV, DSR and OLSR routing algorithm and information to be retransmitted, possibly causing severe delays
interface queue length 50 at each node. The source nodes are in the overall transmission. Drop can be defined as:
respectively 6, 15 and 5 and the receiving nodes are
respectively 0, 1 and 11. No of Packets Dropped = No of pkt Sent – No of pkt Received

TABLE II. NUMBERS OF PACKETS DROP FOR NEW-RENO

Figure 1. Simulation Topology in NS2 environment

https://www.coursehero.com/file/52872281/Performance-comparison-of-TCP-variants-opdf/
 Total Total Total
Routing
Packet Type Sent Received dropped
Protocol
packets packets packets
AODV 2895 2823 72
DSDV 2699 2641 28
New-Reno DSR 3053 2977 76
OLSR 2941 2876 65

This study source was downloaded by 100000824502131 from CourseHero.com on 05-05-2021 13:15:58 GMT -05:00
TABLE III. NUMBERS OF PACKETS DROP FOR RENO TABLE VIII. DATA FOR THROUGHPUT IN TAHOE
Total Total Total Total
Routing Total Sending
Packet Type Sent Received dropped Routing Receiving
Protocol Packet Type Throughput
packets packets packets Protocol Throughput
AODV 3016 2971 45 (kbps)
(kbps)
DSDV 3132 3071 61 AODV 4164192 2428048
Reno DSR 3042 2973 69 DSDV 4549560 2251976
OLSR 2694 2625 69 Tahoe DSR 4326000 2367488
OLSR 4131408 2394000
TABLE IV. NUMBERS OF PACKETS DROP FOR TAHOE
TABLE IX. DATA FOR THROUGHPUT IN VEGAS
Total Total Total
Routing
Packet Type Sent Received dropped Total
Protocol Total Sending
packets packets packets Routing Receiving
AODV 3135 3085 50 Packet Type Throughput
Protocol Throughput
DSDV 2706 2644 62 (kbps)
Tahoe (kbps)
DSR 3110 3032 68
AODV 4131840 2531256
OLSR 2645 2585 60
DSDV 4564992 2343520
Vegas DSR 3911680 2467840
TABLE V. NUMBERS OF PACKETS DROP FOR VEGAS OLSR 3970048 2497208
Total Total Total C. Delay
Routing
Packet Type Sent Received dropped
Protocol
packets packets packets A specific packet is transmitting from source to destination
AODV 2558 2546 12 and calculates the difference between send times and received
DSDV 3283 3257 26 times. Delays due to route discovery, queuing, propagation
Vegas DSR 2153 2146 7
OLSR 2416 2394 22 and transfer time are included in the delay metric. Delay can
be defined as:
B. Throughput
Packet Delay = packets receive time – packet send time
Throughput is the measurement of number of packets
passing through the network in a unit of time. This metric
shows the total number of packets that have been successfully
delivered to the destination nodes and throughput improves
with increasing nodes density. Throughput can be defined as:
Σ Node Throughputs of Data
Transmission Total number of
nodes

TABLE VI. DATA FOR THROUGHPUT IN NEW-RENO

Total
Total Sending
Routing Receiving Figure 2. Delay for TCP New Reno for 100 sec
Packet Type Throughput
Protocol Throughput
(kbps)
(kbps)
AODV 859424 449540
DSDV 4533600 2264764
New-Reno DSR 4145040 2386468
OLSR 4129280 2414216

TABLE VII. DATA FOR THROUGHPUT IN RENO

Total
Total Sending
Routing Receiving
Packet Type Throughput
Protocol Throughput
(kbps) (kbps)
AODV 4076944 2501464
DSDV 4533600 2244016
Reno DSR 4172704 2399304 Figure 3. Delay for TCP Reno for 100 sec
OLSR 4129280 2424856

https://www.coursehero.com/file/52872281/Performance-comparison-of-TCP-variants-opdf/
This study source was downloaded by 100000824502131 from CourseHero.com on 05-05-2021 13:15:58 GMT -05:00

https://www.coursehero.com/file/52872281/Performance-comparison-of-TCP-variants-opdf/
 Figure 4. Delay for TCP Tahoe for 100 sec Figure 8. Jitter for TCP Tahoe for 100 sec

Figure 5. Delay for TCP Vegas for 100 sec Figure 9. Jitter for TCP Vegas for 100 sec

D. Jitter VIII. RESULT ANALYSI TABLE

Jitter is the variation of the packet arrival time. In jitter
calculation the variation in the packet arrival time is expected TABLE X. TOTAL RECEIVING THROUGHPUT RATE, PKT DELIVERY AND
to minimum. The delays between the different packets need to DROP RATE FOR TCP NEW-RENO
be low if we want better performance in Mobile Ad-hoc Total
Drop
Networks. Jitter can be defined as: Packet Type
Routing Receiving
Rate
Delivery
Protocol Throughput Rate (%)
(%)
Jitter ( i ) = Delay (i+1) – Delay (i) where i = 1,2,3…..n Rate (%)
AODV 52.31 % 2.49 % 97.51 %
DSDV 49.96 % 1.04 % 97.85 %
New-Reno DSR 57.57 % 2.49 % 97.51 %
OLSR 58.47 % 2.21 % 97.79 %

TABLE XI. TOTAL RECEIVING THROUGHPUT RATE, PKT DELIVERY AND

DROP RATE FOR TCP RENO
Total
Drop
Routing Receiving Delivery
Packet Type Rate
Protocol Throughput Rate (%)
Rate (%) (%)
AODV 61.36 % 1.49 % 98.51 %
DSDV 49.50 % 1.95 % 98.05 %
Reno DSR 57.50 % 2.27 % 97.73 %
Figure 6. Jitter for TCP New Reno for 100 sec OLSR 58.72 % 2.56 % 97.44 %

TABLE XII. TOTAL RECEIVING THROUGHPUT RATE, PKT DELIVERY AND

DROP RATE FOR TCP TAHOE

Figure 7. Jitter for TCP Reno for 100 sec

 Total
Drop
Routing Receiving Delivery
Packet Type Rate
Protocol Throughput Rate (%)
Rate (%) (%)
AODV 58.31 % 1.59 % 98.41 %
DSDV 49.50 % 2.29 % 97.71 %
Tahoe DSR 54.73 % 2.19 % 97.49 %
OLSR 57.95 % 2.29 % 97.73 %
This study source was downloaded by 100000824502131 from CourseHero.com on 05-05-2021 13:15:58 GMT -05:00