shikh
TRANSCRIPT
-
7/30/2019 shikh
1/20
Abstract
This document is mainly about the influence of protocols on the performance of Satellite
Communication. It is widely recognized that satellite communications are affected by some
peculiar problems, which penalize heavily the performance and efficiency of the
Transmission Control Protocol (TCP). In fact, wireless satellite channels are usually
characterized by link-asymmetry and higher Round Trip Time and Bit Error Rate in
comparison to wired links. It means that TCP, which was developed for wired channels,
exhibits often poor performance in a satellite scenario (unfair bandwidth repartition, low
throughput and long file-transfer delay). We will study and compare many TCP variants
recently proposed. In particular we will compare TCP-Reno standard implementation with
TCP-SACK, TCP-Westwood, TCP-Vegas and TCP-Tibet. We will test the different
protocols on a simulated satellite scenario, with the support of NS2, a well-know network
simulator platform, annotating the advantages and drawbacks of various protocols in order
to improve the transmission of IP data packets over satellite channels.
Keywords: Satellite link, Performance analysis, TCP behaviors.
-
7/30/2019 shikh
2/20
Introduction
The use of satellites for Internet traffic is a very attractive proposition since the wired
network is highly congested. On the average, any IP flow needs to travel through 17 to 20
nodes, and hence may go across multiple bottlenecks. On the other hand, with satellites, just
one hop is sufficient to connect two very distant sites.
Unfortunately, the use of satellite is limited by the
poor performance obtained by usual TCP/IP suite protocols. In fact the long propagation
delay of satellite channel has the same effect of traveling through various bottlenecks
affecting the congestion control scheme of many connection oriented protocols, which
erroneously underestimate the bandwidth available on the link and assume to be in a
congestion state. Moreover, in the future communication system, the satellite component is
supposed to be integrated with the terrestrial component [1] and the interaction of satellite
and terrestrial connections brings new performance problems which are actually partially
unexplored and not completely solved. Satellite communication needs to be completely
integrated in the actual communication system without technical handicaps, if it wants to
assume a particularly important role, and not only complementary, in the future
communication world, especially if aiming at real global coverage and ensuring access to
maritime, aeronautical and remote users.
The main goal of this project is to evaluate the performance
of typical mobile Internet applications, using different TCP schemes. While high efficiency
on fast satellite links may eventually require the introduction of a new protocol, many
studies [2] indicate that full link utilization may be achieved by using the TCP performance
enhancements that have already been approved or that are currently in the standards process.
Of particular interest in the satellite environment are performance enhancements for scaled
windows and timestamps, fast retransmit and recovery, and selective acknowledgement.
This paper describes the protocols which adopt these performance enhancements and show
the effects of their implementation on increasing TCPs performance oversatellite links.
-
7/30/2019 shikh
3/20
We analyze some representative satellite scenarios with a
classical GEO configuration satellite repeater.
-
7/30/2019 shikh
4/20
Literature survey
THE ROLE OF SATELLITES:
In the future, satellites will play an important role in providing wireless access to mobile
users, for enhancing coverage of existing or developing land mobile systems and to ensure
long-range mobility and large bandwidth. The presence of satellite links provides extra
capacity and an alternate and less congested route to existing wired and wireless systems,
thus offering unique opportunities for improving efficiency and reliability. The broadcast
nature of the satellite channel also makes satellites suitable for broadcasting/multicasting.
Besides the satellite is the only way for extending the range of communication out of our
planet connecting space stations and space shuttles in a future interplanetary communication
network. The growing interest in exploring new worlds needs an affordable and performing
system of communication for granting the success of space missions, in order to guarantee a
complete control of mission evolution.
The satellite may also be integrated with short-range wirelessnetworks (Bluetooth) in order to provide Internet services to mobile vehicles. In this
scenario, the satellite will connect one vehicular terminal to the Internet while the Bluetooth
system will be able to interconnect equipment inside the vehicle (car, bus, train, plane or
ship).
The main satellite communication systems, providing large bandwidth
directly to users, are based on a geosynchronous orbital configuration. A single GEOsatellite is able to cover a very wide area with multiple narrow spot beams using multibeam
antennas. In this way, very reliable links can be provided even with small terminals. Such
GEO systems provide (or are designed to provide) high availability high data rate links (up
to 2 Mbit/s and more) and huge total capacity (of the order of 10 Gbit/s). Many of them are
expected to provide regional limited coverage in a few years and global coverage
(excluding poles) finally.
-
7/30/2019 shikh
5/20
Previous works [8-12] has addressed the performance of TCP/IP over
geostationary satellites. This has mainly focused on evaluating performance of different
TCP schemes (Tahoe, Reno, New Reno, SACK), where the connection between two fixed
stations goes across a GEO satellite. In this paper, we also address other TCP schemes
(Westwood [12] and Tibet[7]) and evaluate TCP performance when terrestrial and satellite
TCP connection share the same bottleneck in order to establish which protocol has the
better behaviour when used for both long and short Round Trip Time (RTT) connection;
besides we exposed both type of connections to the aggressive behavior of a UDP flow to
evaluate the ability to conserve their throughput. Such an environment creates a very
realistic scenario for the evaluation of performance in a satellite environment.
SATELLITE FEATURES AFFECTING PERFORMANCE OF IP APPLICATIONS
First, when considering the performance of real time (e.g.
voice and video) applications the large propagation delays represent a critical issue. The
problem is particularly acute in TCP applications over links with large bandwidth-delay
products, where a large TCP window is required for efficient use of the satellite link. A
lossy satellite link can cause frequent slow start events, with significant impact on
throughput. The key satellite network features that need to be considered in order to
evaluate the impact of satellite on internet application performance are: propagation delay,
Delay- Bandwidth Product (DBP), signal-to-noise ratio (SNR), satellite diversity, routing
strategy.
Satellite systems are changing their role from the bent-
pipe (transparent) channel paradigm to the onboard routing (regenerative) paradigm
associated with packet transmission and switching. In this process each satellite is anelement of a network, and new design problems are involved both at the network and the
transport layers.
While considering the performance of standard Internet protocols over paths including
satellite links, the large propagation delays have to be taken into account. The problem is
particularly relevant to TCP due to its delay-sensitive nature. For voice and other real time
applications too, the delay may be critical. Some of the main aspects that need to be
-
7/30/2019 shikh
6/20
considered in order to evaluate the impact of system architecture on the performance of
network protocols are:
Propagation delay:
Due to the large distance that packets need to travel in satellite networks, they can
experience a significant transmission delay (depending on the type of constellation and the
routing strategy adopted). The transmission delay in a GEO architecture depends on the
user-gateway distance (or user-user if direct connection is allowed) when a single satellite
configuration with no ISLs (inter-satellite links) is considered, or on the connection strategyif ISLs are used. This delay can be considered constant in the caseof fixed users while it is
variable in the case of mobile users. However due to queue phenomenon in terrestrial router
even fixed users can experience a large variability in propagation delay. This phenomenon
is even more critical if ISLs are implemented, since the routing strategy may also play a
very important role. The delay variability is particularly relevant to TCP, since it causes the
delivery packet time to be variable and may affect the estimate of Round Trip Time (RTT)
by TCP. Whenever a change in RTT occurs, it takes some time for TCP to adapt to the
change. If the RTT keeps changing, TCP may not be able to update its estimate of RTTquickly enough. This may cause premature timeouts/retransmissions, reducing the overall
bandwidth efficiency.
Delay-Bandwidth Product (DBP):
The performance of TCP over satellite links is also influenced by the Delay-Bandwidth-
Product (DBP). This has been evaluated through simulations and experiments [9, 10, and11]. This performance can be enhanced by adopting techniques such as selective
acknowledge and window dimensioning [9, 11]. Also, techniques such as TCP spoofing,
splitting and cascading may be used [11].
Satellite diversity:
Satellite diversity is a very efficient technique for improving link availability or SNR byutilizing more than one user-satellite link to establish communication. In a connectionless
-
7/30/2019 shikh
7/20
system, the satellite diversity feature has the advantage of increasing the probability of
packet delivery to/from the satellite network from/to Earth. However, this increases the
number of busy channels, which causes a reduction in capacity.
Signal-to-noise ratio (SNR):
In GEO constellations, the SNR (or equivalently Bit Error Rate, BER) is characterized by a
great variability due to free space losses and troposphere propagation (including rain, over
10 GHz) for fixed communications and also due to shadowing in case of mobile
communications. Both shadowing and deep rain fading can cause packet loss. Poor SNR is
an extensively studied issue for GEO satellites. Since TCP has been designed mainly for
wired links, it assumes that the BER is very low (of the order of 10-10, or less) so that when
a packet is lost, TCP attributes it to network congestion. This causes TCP to employ its
congestion control algorithms, which reduces the overall throughput. This is detrimental in
satellite links, and causes an unnecessary reduction in throughput.
4. TCP PROTOCOLS DETAILS
The Transmission Control Protocol (TCP) was originally designed for wired link so it is
based on some assumption related to specific characteristic of wired link. Moreover TCP is
based on the assumption that the network does not provide any explicit feedback to the
sources. Therefore each source must form its own estimates of the network properties, such
as round-trip time (RTT) or usable bandwidth in order to perform efficient end-to-end
congestion control.
In order to improve the performance of TCP connections for
preventing congestion events and achieve a fair share of bandwidth among different
connections, the original TCP protocol was extended with some additive function, the mains
of which were implemented in TCP Reno. In the following we describe these functions:
-
7/30/2019 shikh
8/20
Slow Start and Congestion Avoidance
The slow start and congestion avoidance mechanisms were introduced in 1988 in [13] and
added as a requirement for TCP implementations in RFC 1122[14].
Slow start is used to gradually increase the rate at which the
sender injects data into the network. Slow start begins by sending one segment and waiting
for an acknowledgement. For each acknowledgement the sender receives, it injects two
segments into the network, thus leading to an exponential increase in the amount of data
being sent. Slow start ends when the receivers advertised window is reached or when loss
is detected. Because the amount of time required for slow start to achieve full bandwidth is
a function of round trip time, satellite links are particularly sensitive to the limited
throughput available during slow start.
Congestion avoidance is used to probe the network
for available bandwidth by sending one additional segment for each round trip time (up to
the receivers advertised window). In the original slow start/congestion avoidance scheme,
when the sending TCP detects segment loss (by default assuming network congestion), itdrops back into slow start until the packet sending rate is one half the rate at which the loss
was detected and then begins the congestion avoidance phase.
Fast Retransmit and Fast Recovery
TCP Reno also incorporates Fast Retransmit and Recovery, RFC 2001[15]. Although these
mechanisms have been found in many Unix variants of TCP for several years, they were not
documented as a Standards Track RFC until 1997. Fast retransmit reduces the time it takes a
TCP sender to detect a single dropped segment. Rather than waiting for the retransmit
timeout (RTO), the TCP sender can retransmit a segment if it receives three duplicate ACKs
for the segment sent immediately before the lost segment.
Fast recovery works in conjunction with fast retransmit. As mentioned above, when a
sender retransmits a segment, it normally recovers by moving first into a slow start phase
followed by a congestion avoidance phase. If the sending TCP detects the segment lossusing fast retransmit, however, fast recovery is used instead. Fast recovery halves the
-
7/30/2019 shikh
9/20
segment sending rate and begins congestion avoidance immediately, without falling back to
slow start.
TCP SACK
This protocol introduces the ability for TCP to use Selective Acknowledgments. This aspect
is considered very important in satellite scenario, because the cumulative positive
acknowledgments employed by TCP are not particularly well suited to the long-delay
satellite environment due to the time it takes to obtain information about segment loss.Using a selective acknowledgement (SACK) mechanism, as defined in RFC 2018[4], the
SACKs generated at the receiver explicitly inform the sender about which segments have
arrived and which may have been lost, giving the sender more information about which
segments might need to be retransmitted and avoiding to retransmit segments which travel
trough long delay link, such as satellite connections.
TCP Newreno
This protocol implements some modifications in Fast Recovery algorithm, in order to
improve the response to partial acknowledgments, which were first proposed by Janey Hoe
[6]. In the original implementation of the TCP Fast Recovery algorithm the TCP data sender
only retransmits a packet after a retransmit timeout has occurred, or after three duplicate
acknowledgements have arrived, triggering the Fast Retransmit algorithm. A single
retransmit timeout might result in the retransmission of several data packets, but each
invocation of the Reno Fast Retransmit algorithm leads to the retransmission of only a
single data packet. TCP NewReno defines a "Fast Recovery procedure" that begins when
three duplicate ACKs are received and ends when either a retransmission timeout occurs or
an ACK arrives that acknowledges all of the data up to and including the data that was
outstanding when the Fast Recovery procedure began. In this way it can recover a situation
when multiple packets have been dropped from a single window of data in a faster way than
TCP-Reno.
TCP Vegas
In order to increase throughput and decrease losses TCP Vegas introduces three techniques[18]. Firstly, a more timely retransmission of dropped segments is used. This is obtained by
-
7/30/2019 shikh
10/20
-
7/30/2019 shikh
11/20
BACKGROUND
1. Reactive Routing Protocol
These protocols only find a route to the destination node when there is a need to send data.
The source node will start by transmitting route request throughput the network. The sender
will than wait for the destination node or an intermediate node to respond with a list of
intermediate nodes between the source and destination.
Reactive protocols start to set up routes on-demand. The routing protocol will try to
establish such a route, whenever any node wants to initiate communication with another
node to which it has no route. This kind of protocols is usually based on flooding the
network with Route Request (RREQ) and Route reply (RERP) messages.
Reactive routing protocols are:
protocol (DSR)
-Based Adaptive Routing (SSA)
-Aided Routing Protocol (LAR)
-
7/30/2019 shikh
12/20
2. Proactive Routing Protocol
Proactive WSNs protocols are also called as table-driven protocols and will actively
determine the layout of the network. Through a regular exchange of network topology
packets between the nodes of the network, at every single node an absolute picture of the
network is maintained. There is hence minimal delay in determining the route to be taken.
This is especially important for time-critical traffic. When the routing information becomes
worthless quickly, there are many short-lived routes that are being determined and not usedbefore they turn invalid.
Proactive WSN Protocols include:
-eye State Routing (FSR)
-Sequenced Distance Vector (DSDV)
-head Gateway Switch Routing Protocol (CGSR)
3. Hybrid Routing Protocol
Since proactive and reactive protocols each work best in oppositely different scenarios,
hybrid method uses both. It is used to find a balance between both protocols. Proactive
operations are restricted to small domain, whereas, reactive protocols are used for locating
nodes outside those domains.
Hybrid protocols are:
ting Protocol (ZRP) 13
-
7/30/2019 shikh
13/20
-head Gateway Switch Routing Protocol (CGSR)
The routing protocols AODV, DSR and DSDV are three of the promising routing protocols.
They can be used in mobile ad hoc networks to rout packets between mobile nodes.
Problem formulation and implementation issues
TCP/IP protocol was designed for wired networks which provides end to end reliable
communication between nodes and assures ordered delivery of packets. It also provides
flow control and error control mechanisms. As it is still a successful protocol in wired
networks, losses are mainly due to congestion. But in case of ad hoc networks packet losses
are due to congestion in the network and due to frequent link failures so when we adapt
TCP to ad hoc networks it misinterprets the packet losses due to link failure as packet losses
due to congestion and in the instance of a timeout, backing-off its retransmission timeout
(RTO).This results in unnecessary reduction of transmission rate because of which
throughput of the whole network degrades. Therefore, route changes due to host mobility
can have a detrimental impact on performance. On-demand Routing Protocols such as
AODV and DSR are used for this performance analysis of TCP. These types of routing
protocols create routes only when requested by a source node. When a node wants to
establish a route to a destination, it initiates a route discovery process within the network.
One the route has been established, it is maintained until either destination becomes
inaccessible or the route is no longer desired. We will analyze the performance of two on-
demand routing protocols for ad hoc networks, namely Dynamic Source Routing (DSR) and
Ad Hoc On-demand Distance Vector (AODV) routing protocols, based on TCP traffic
flows. We also use DSDV which is a table driven routing protocol.
-
7/30/2019 shikh
14/20
Proposed research methodology
TCP optimization in WSNs has been investigated in several studies. TCP does not have any
resilience mechanisms that are specially designed to deal with link failures. From the
viewpoint of TCP, there is no difference between link failure and network congestion. As a
result, when part of the network fails and some segments are dropped, TCP will assume that
there is congestion somewhere in the network, and the TCP congestion control mechanisms
will start dealing with the segment loss. TCP congestion control mechanisms have
improved over time. The main versions of TCP are Tahoe TCP, Reno TCP, NewReno TCP
and SACK TCP. Tahoe TCP is the oldest version and only a few old systems use it.
NewReno TCP and SACK TCP are widely implemented. We focus on another two new
TCP variants TCP FACK and TCP Vegas because they are the newer versions and are
being deployed. And also the analysis was not on just one routing protocol but three widely
used routing protocols, AODV, DSR and DSDV Congestion Control Mechanisms The
predominant example of end-to-end congestion control in use today, that implemented by
TCP. The essential strategy of TCP is to send packets into the network without a reservation
and then to react to observable events that occur. TCP assumes only FIFO queuing in the
network routers, but also works with fair queuing. Fast Retransmit and Fast Recovery themechanisms described so far were part of the original proposal to add
congestion control to TCP. It was soon discovered, however, that the coarse-grained
implementation of TCP timeouts led to long periods of time during which the connection
went dead while waiting for a timer to expire. Because of this, a new mechanism called fast
re- transmit was added to TCP. Fast retransmit is a heuristic that sometimes triggers the
retransmission of a dropped packet sooner than the regular timeout mechanism.
-
7/30/2019 shikh
15/20
Required resource for proposed work
Network Simulator (NS-2)
Ns2 is a discrete event simulator targeted at networking research. It provides substantial
support for simulation of TCP, routing and multicast protocols over wired and wireless
networks. It consists of two simulation tools. The network simulator (ns) contains all
commonly used IP protocols. The network animator (nam) is use to visualize the
simulations. Ns2 fully simulates a layered network from the physical radio transmission
channel to high-level applications.Ns2 is an object-oriented simulator written in C++ and
TCL. The simulator supports a class hierarchy in C++ and a similar class hierarchy within
the TCL interpreter. There is a one-to-one correspondence between a class in the interpreted
hierarchy and one in the compile hierarchy. The reason to use two different programming
languages is that OTCL is suitable for the programs and configurations that demand
frequent and fast change while C++ is suitable for the programs that have high demand in
speed. Ns2 is highly extensible. It not only supports most commonly used IP protocols but
also allows the users to extend or implement their own protocols. The latest ns2 versionsupports the four routing protocols, including DSR. It also provides powerful trace
functionalities, which are very important in our project since various information need to be
logged for analysis. The full source code of ns2 can be downloaded and compiled for
multiple platforms such as UNIX, Windows.
Languages Used:-
1. C++
-
7/30/2019 shikh
16/20
C++ is a programming language that implements object-oriented programming. C++ uses a
compiler, each time something in the source code is changed, the program has to be
partially recompiled and delinked. C++ is a superset of C, which means that a programmer
is free to use C code for the speed critical parts of a program.
1. TCLTCL is an interpretive language. In TCL, a programmer can add new commands to the
language by implementing them as C functions. The C functions can then be called from the
(8)
command line interface of the TCL interpreter.
7. Expected outcomes
In this article By using different WSN routing protocols like proactive (DSDV) and reactive
(AODV and DSR) we want to measure, calculate and compare the parameters like
Throughput, End To End Delay and Packet Delivery Ratio.
1. Throughput: Throughput is the average number of successfully delivered datapackets on a communication network or network node.
Throughput is calculated in bites/sec, data packet/second and data packet/time slot.
Total number of received packets at destination
Throughput (bites/sec) = -------------------------------------------------------------
Total simulation time
2. End to end Delay: It can be defined as the time a packet takes to travel fromsource
to destination.
-
7/30/2019 shikh
17/20
dend-end= N[ dtrans+dprop+dproc]
Where
dend-end = end-to-end delaydtras = transmission delay
dprop = propagation delay
dproc = processing delay
N = number of links (Number of routers + 1)
The lower value of end to end delay means better performance of the protocol.
3.
Packet Delivery Ratio: The ratio of number of delivered data packet to the(9)
destination.
Total number of packet receive
PDR = -------------------------------------------
Total number of packet sent
The greater value of the packet delivery ratio means better performance of the
Protocol.
By using this fundamentals we are comparing all the rounting protocol which are taken
by us. We are expecting that after the comparison reactive (AODV) routing protocol is
much better than other two.
http://en.wikipedia.org/wiki/Propagation_delayhttp://en.wikipedia.org/wiki/Processing_delayhttp://en.wikipedia.org/wiki/Processing_delayhttp://en.wikipedia.org/wiki/Propagation_delay -
7/30/2019 shikh
18/20
References
1. D. OMahony, UMTS: the fusion of fixed and mobile networking, IEEE Internet
Computing, vol. 21, Jan.-Feb. 1998.
2. M. Allman, C. Hayes, H. Kruse, S. Ostermann, TCP Performance over Satellite Links,Ohio University
3. P. Loreti, M.Luglio, R. Kapoor, J. Stepahnek, M. Gerla, F. Vatalaro, and M. A. Vazquez-
Castro, Satellite Systems Performance with TCP-IP Applications, IWDC 2001, Taormina,
September 2001.
4. S. Floyd, M. Mathis, J.Mahadavi, and A.Romanov, TCP Selective Acknoledgement
Option, RFC 2018, April 1996
5. S. Floyd, and T.R.Henderson, The NewReno Modifications to TCPs Fast Recovery
Algorithm, IETF RFC 2582, April 1999
6. J.C: Hoe, Improving the Start-up Behaviour of a congestion Control Scheme for TCP,
ACM SIGCOMM Computer communications Review, October 1996.
7. A. Capone, and F. Martignon, Bandwidth Estimates in the TCP Congestion Control
Scheme, IWDC 2001, Taormina, September 2001.
-
7/30/2019 shikh
19/20
8. W. D. Ivancic, D. Brooks, B. Frantz, D. Hoder, D. Shell, D. Beering, NASAs Broadband
satellite networking research, IEEE Communications Magazine, n. 7, July 1999, pp. 40-47.
9. C. P. Charalambous, V. S. Frost, J. B. Evans, Performance evaluation of TCP extensionson ATM over high bandwidth delay product networks, IEEE Communications Magazine, n.
7, July 1999, pp. 57-63.
10. H. Kruse, S. Ostermann, M. Allman, On the Performance of TCP-based Data transfers
on a faded Ka-Band Satellite Link, Proceedings of the 6th Ka-Band Utilization Conference,
June 2000.
11. M. Allman, H. Kruse, S. Ostermann, A History of the Improvement of Internet
Protocols Over Satellites Using ACTS, Invited paper for ACTS Conference 2000, May
2000.
12. C. Casetti, M. Gerla, S. S. Lee, S. Mascolo, M. Sanadidi, TCP with Faster Recovery,
UCLA CSTechnical Report #200017, 2000
13. V. Jacobson, Congestion Avoidance and Control, In Proceedings ACM SIG-COMM 88,
ACM, August 1988
14. R. Braden, Requirements for Internet Hosts- Communication Layers, October 1989,
RFC 1122
15. W. Stevens, TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast
Recovery Algorithms, January 1997, RFC 2001
16. J. Postel, Transmission Control Protocol, September 1981, RFC 793
17. V. Jacobson, R. Braden, and D. Borman, TCP Extensions for High Performance, May
1992, RFC 1323
18. L. S. Brakmo, S. W. OMalley, L. L. Peterson, TCP Vegas: New Techniques forCongestion Detection and Avoidance, 1994 ACM SIGCOMM Conference, pages 24--35,
May 1994.
19. Network Simulator (NS-2), www.isi.edu/nsnam/ns/.
-
7/30/2019 shikh
20/20