Simulation Results

woonsocketpoliticalΔίκτυα και Επικοινωνίες

28 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

76 εμφανίσεις

1
University of Massachusetts, DartmouthUniversity of Massachusetts, Dartmouth
Computer and Information Science DepartmentComputer and Information Science Department
Supervisor: Dr. Emad Aboelela
Student: Phuong Tu
Spring-2002
Master ProjectMaster Project
::
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Abstract
Abstract
 The purpose of this project is to model and analyze network
performances by using OPNET as a simulation tool. The
project covers 3 chapters:
 Multiple Access Protocols: Compare the performance
between ALOHO and CSMA/CD network protocols.
 Routing Protocols: Models and analyze several problems in
routing protocols including RIP, IGRP, EIGRP and OSPF.
 TCP-Transmission Control Protocol: Examine Slow Start and
Congestion Control Algorithm.
2
Chapter 0
Chapter 0
OPNET
INTRODUCTION
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
What is the OPNET Modeler?
What is the OPNET Modeler?
OPNET is a comprehensive software environment for
• Network modeling and simulating.
• Analyzing the performance of communications networks, computer
systems and applications, and distributed systems.
OPNET provides a well-built and detailed graphical user interface
OPNET is also used as a decision support tool to provide insight
into the performance and behavior of existing or proposed networks,
systems, and processes.
3
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Three
Three
-
-
Tiered OPNET Hierarchy
Tiered OPNET Hierarchy
 Three domains: network, node, and process
 Network models specify network topology
 Node models specify objects in network domain
 Process models specify objects in node domain
Chapter 1: MULTIPLE ACCESS PROTOCOL
Chapter 1: MULTIPLE ACCESS PROTOCOL
ANALYZE PERFORMANCE OF
ALOHA AND CSMA/CD PROTOCOLS
4
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
-
-
ALOHA
ALOHA
 Pure ALOHA:
Aloha is the Ethernet that has its roots in an early packet radio network.
Stations are allowed access to the channel whenever they have data to
transmit. To avoid data collision, each station must either monitor its
transmission on the rebroadcast or await an acknowledgment from the
destination station.
 Performance of ALOHA
 Throughput
:
 S: number packets successfully (without collision) transmitted per
unit time
 Offered load: G: number packets transmissions attempted per unit
time, S<G, but S depends on G
S = Ge-2G
 Aloha throughput
 at most 18% of bandwidth
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
CSMA/CD
CSMA/CD
 Ethernet uses a refinement of ALOHA, known as Carrier Sense
Multiple Access (CSMA), which improves performance when there
is a higher medium utilization. With Collision Detection, each
transmitting node monitors its own transmission, and if it observes a
collision, it stops transmission immediately and instead transmits a
32-bit jam sequence.
 The CSMA/CD protocol is designed to provide fair access to the
shared channel so that all stations get a chance to use the network.
After every packet transmission all stations use the CSMA/CD
protocol to determine which station gets to use the Ethernet channel
next.
5
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Modeling and Simulation
Modeling and Simulation
 The ALOHA model consists of generic
transmitter nodes and one receiver node.
 Node 0 is Generic Receiver Node.
Others are generic transmitter nodes.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Aloha: Network Topology
Aloha: Network Topology
6
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
CSMA/CD: Network Topology
CSMA/CD: Network Topology
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Configuration
Simulation Configuration
 Goal: Observe how the performance of the ALOHA
and CSMA/CD protocols vary as a function of the
channel traffic.
 Parameters are set for this simulation.
 Packet Size: 1024 bits
 Inter-arrival: 1000, 200, 150, 100, 80, 50, 35, 30, 25, 20,
18, 5.
 Data Rate: 1024 bits/sec
 Packet transmission time t = 1.0s
 Max Packet count = 1000. In that sense, each node will
generates about 50 packets.
7
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Aloha: Simulation Results
Aloha: Simulation Results
• Simulation Result:
• Theoretical Aloha throughput
• S = Ge
-2G
Theoretical ALOHA Channel Throughput vs Channel Traffic
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Channel Traffic G
Channel Throughput S
The experiment shows that the result is very close to the theoretical ones. The
maximum throughput is achieved near G = 0.5 and Smax ~ 0.18.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
CSMA/CD: Simulation Results
CSMA/CD: Simulation Results
• Simulation Result:
• Theoretical Aloha throughput
• S = Ge
-2G
The experiment shows that the result is very close to the theoretical ones.
Theoretical CSMA/CD Channel Throughput vs Channel Traffic
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Traffic G
Throughput S
8
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Conclusion
Conclusion
 CSMA/CD outperforms ALOHA Protocol in
the term of ETE-delay and throughput as
CSMA/CD adds a carrier sense capability to
sense the channel and determine that it is free
before committing to a transmission. This
helps the network to detect data collision.
Chapter 2.1: ROUTING PROTOCOL
Chapter 2.1: ROUTING PROTOCOL
Analyze and Simulate Routing Protocols
Analyze and Simulate Routing Protocols
OSPF, RIP, IGRP and EIGR
OSPF, RIP, IGRP and EIGR
9
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
 RIP: The Routing Information Protocol, or RIP, as it is more
commonly called, is one of the most enduring of all routing
protocols. RIP use distance vectors to mathematically compare
routes to identify the best path to any given destination
address. As a distance-vector based algorithm, RIP works fine
for small, stable high-speed networks.
 IGRP: is a distance vector Interior Gateway Protocol IGRP
uses a composite metric that is calculated by factoring
weighted mathematical values for internet work delay,
bandwidth, reliability, and load. To provide additional
flexibility, IGRP permits multi-path routing
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
 EIGRP: The Enhanced Interior Gateway Routing Protocol
represents an evolution from its predecessor IGRP. This evolution
resulted from changes in networking and the demands of diverse,
large-scale internet works. Enhanced IGRP integrates the
capabilities of link-state protocols into distance vector protocols. It
consists of Diffusing update algorithm (DUAL)
 OSPF:is a link-state routing protocol that calls for the sending of
link-state advertisements (LSAs) to all other routers within the same
hierarchical area. As a link-state routing protocol, OSPF contrasts
with RIP and IGRP, which are distance-vector routing protocols.
OSPF addresses all RIP shortcomings and thus is better suited for
modern large, dynamic networks
10
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Case Study
Case Study
 Implement a network that consists of 50 subnets and
an Internet Cloud. The network support Email
application. There are 4 simulation scenarios. Each
implements the routing protocol of RIP, IGRP,
EIGRP and OSPF.
 We then observe how each routing protocol generates
traffic in whole network and also observe network
perform in term of throughput, utilization and queue
delay.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Network Topology
Network Topology
11
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
031,305,056491,185OSPF Traffic Sent (bits/sec)
41,98422,455,627270,653EIGRP Traffic Sent (bits/sec)
39,42422,442,997267,945 EIGRP Traffic Received
(bits/sec)
0.00160.03940.0268EIGRP Convergence Time (sec)
3423,530,9801,022,038IGRP Traffic Sent (bits/sec)
922,187,139917,771IGRP Traffic Received (bits/sec)
013,903,5522,518,191RIP Traffic Sent (bits/sec)
013,483,2802,370,206RIP Traffic Received (bits/sec)
MinimumMaximumAverageStatistic
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
 RIP: Routing table updated
every 30s
 IGRP: Routing table updated
every 86s
12
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
 EIGRP: Routing table updated
every 30s. But, the traffic is very
small in comparison with others.
 OSPF: Peak occurred at initial
state where flooding occurs.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Compare Results
Compare Results
 Utilization: As Is. RIP give the
heavies traffic in the network.
 Average Utilization. Utilization in
RIP case is almost 4 times higher
than others.
13
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Compare Results
Compare Results
 Throughput: Max throughput
occurred at the time 4.5’ of simulation
time. RIP sends most packets and
then OSPF.
 Average PTP Delay: Except RIP,
other routing protocols have almost
PTP delay. The delay occurred in RIP
is so high as RIP needs to update
routing table every 30’.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Conclusion
Conclusion
 From the simulation results, we could observe that
EIGRP has the less traffic in the network and RIP
generates the heaviest one. It is because RIP arranges
to have routers to broadcast their entire current
routing database periodically, typically every 30
seconds. In contrast to a distance-vector algorithm,
where a router ``tells all neighbors about the world,''
link-state routers ``tell the world about the
neighbors'', that reduces significantly the traffic in the
networks.
14
Chapter 2.2: ROUTING PROTOCOL
Chapter 2.2: ROUTING PROTOCOL
Analyze and Simulate
Analyze and Simulate
RIP Routing Protocols
RIP Routing Protocols
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
 RIP is a distance-vector protocol that allows routers to exchange information about
destinations for computing routes throughout the network. RIP is used primarily in
homogeneous networks of moderate size. Distance-vector algorithms make each
router periodically broadcast its routing tables to all its neighbors. Then a router
knowing its neighbors' tables can decide which destination neighbor to use for
routing a packet. [1]
 RIP version 1 (RFC 1058): [2]
RIPv1 is a simple distance vector protocol. It has been enhanced with various
techniques, including Split Horizon and Poison Reverse in order to enable it to
perform better in somewhat complicated networks.
 The longest path cannot exceed 15 hops.
 RIP uses static metrics to compare routes.
 RIP version 2 (RIPv2): RIP version 2 (RIPv2) adds several new features.
 External route tags.
 Subnet masks.
 Next hop router addresses.
 Authentication.
 Multicast support.
15
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Case Study
Case Study
Design a campus network. It is comprised of four
fully meshed backbone routers that support four
major networks, each with a central router and 2 sub-
networks. Each subnet has 2 sub-network, each of
which support 10 stations.
Observe the performance of this campus network with
2 version RIPv1 and RIPv2
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Network Topology
Network Topology
16
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Assign IP Address
Assign IP Address
150.70.20.0/24150.70.10.0/24150.70.0.0/16South
150.50.20.0/24150.50.10.0/24150.50.0.0/16East
150.30.20.0/24150.30.10.0/24150.30.0.0/16West
150.10.20.0/24150.10.10.0/24150.10.0.0/16North
Subnet EastSubnet West
IP Address
IP AddressCampus
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Configuration
Simulation Configuration
 Simulation Duration: 10 minutes.
 IP Dynamic Protocol: RIP or Default.
 RIP Simulation Efficiency: Disabled.
 Values of statistics: 3600
 RIPv1
 All routers are defined running under RIP version 1. This attribute
is set in Router Attribute as RIP Parameters.
 RIP Parameter: Enable Auto Summary.
 RIPv2
 All routers are defined running under RIP version 2.
 RIP Parameter: Disable Auto Summary.
17
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
 RIPv1  RIPv2
RIPv2 generate more traffic than RIPv1 as we disabled Auto Summary
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
Incoming traffic to East Campus
LAN.
Total number of Broadcast IP
datagram received at East Backbone node.
18
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Conclusion
Conclusion
Network Traffic: RIPv2 generate heavier traffic than RIPv1. It
is because we set Auto Summary is disabled in the RIPv2
scenario. Hence, sub-networks are advertised outside of their
major network. The traffic in each campus is therefore
increased.
Route Table:
 RIPv1: The routing table of one of the Central routers we see how
the subnets for the other major networks are summarized so they
only have an entry for the major networks.
 RIPv2: In contract to RIPv1, we would see the major networks are
not summarized, and the route table contains all of the sub-
networks.
Chapter 2.3: ROUTING PROTOCOL
Chapter 2.3: ROUTING PROTOCOL
Analyze RIP
Analyze RIP
with Triggered Extension Mode
with Triggered Extension Mode
19
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
Triggered Extensions to RIP suggests an enhancement to the
"Routing Internet Protocol" (RIP) and "RIP-2" to allow them
to run more cost-effectively on Wide Area Networks (WANs).
Triggered RIP requires that there is an underlying mechanism
for determining unreachability in a finite predictable period.
The triggered extensions to RIP are particularly appropriate for
WANs where the cost - either financial or packet overhead -
would make periodic transmission of routing (or service
advertising) updates unacceptable.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Case Study
Case Study
This scenario simulates a company with offices in both
Washington D.C. and New York, connected by a WAN
Line. The company uses a single database server located in
the New York office; however the Washington office doesn't
heavily use this server.
Conventional RIP routing can prove to be inefficient in this
type of situation due to RIP's periodic updates. In this type
of scenario the routing update traffic can be the major source
of traffic over the WAN link connecting the Washington and
New York offices.
20
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Case Study
Case Study
Network Application:
The company has 4 departments, two of which are located at each office.
The company supports below applications. All of those are supposed to
have light use.
 Telnet
 Email
 Database
 File Printing
New York Office:
It includes Accounting and Human Resources Department. The office is
installed two servers to support Printing File and Database Applications.
Washington Office:
It consists of two departments: Engineering and Information System that
support Email Application and Remote Login respectively. Suppose that
staff from Washington Office rarely access to the database server located
at New York Office
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Network Topology
Network Topology
21
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
Point-to-point throughput (packets/sec)
Washington  New York.
Link’s throughput has decreased in triggered
ON.
Traffic generated in the network. The traffic is
slightly reduced when triggered mode is on.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
Point-to-point throughput (bits/sec) Washington  New York.
Once Triggered On, the throughput of the link has dramatically decreased.
22
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Conclusion
Conclusion
By observing the simulation results, we could
conclude that when Triggered Extension Mode is on,
the traffic in the network is reduced that also causes
decrease in the throughput of network links.
Triggered Extension Mode is useful when there is a
little change in the WANs network and it would help
to reduce the cost of packet overhead.
Chapter 2.4: ROUTING PROTOCOL
Chapter 2.4: ROUTING PROTOCOL
Analyze impact on Failure Recovery in
Analyze impact on Failure Recovery in
Routing Protocols
Routing Protocols
23
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Case Study
Case Study
 We model a network that consists of 10 subnets located in 20
large cities in USA. The network supports all of basic
applications such as web-browser, email, FTP, Remote login,
Database and so on. The network also controls Failure
Recovery feature by adding the node IP failure recovery node.
It provides attributes for controlling the time and status of
objects in the model.
 We execute 3 scenarios with the same network model. Each
scenario will then conduct routing protocols: RIP, IGRP and
EIGRP. OPNET simulation will give us the performance of
each protocol in dropping packet during link failure.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Network Topology
Network Topology
24
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Configuration
Simulation Configuration
 Simulation Duration: 1,200 sec
 Enable failure Recovery Node.
 Suppose link between New York subnet and internet
cloud:
 Fail at the second 200
 Recover at the second 600
 Set IP Dynamic Protocol for each scenario. The
correspondent protocol SimEfficiency will then be
set Disabled.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
000EIGRP
010.138IGRP
01.080.02RIP
Minimum Maximum Average Statistic
IP. TRAFFIC DROPPED (packets/sec)
25
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
RIP: Drop one packet IGRP: Drop several packets
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
EIGRP: No packet dropped Comparison of packet dropped
26
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
RIP: Generate very heavy traffic IGRP: Less traffic generated
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
EIGRP: Generate very slight traffic Result Comparison
27
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Conclusion
Conclusion
 EIGRP performed the best by not dropping
any traffic; this can be attributed to the DUAL
process which allows EIGRP to poll its
neighbors for routes not currently in its route
table.
 IGRP shows a significant spike which is
affected by the hold-down timer, which
prohibits updates for a removed route.
Chapter 3: TCP
Chapter 3: TCP
Slow start Congestion Avoidance with no drop
Slow start Congestion Avoidance with no drop
.
.
Tahoe network with a default one drop.
Tahoe network with a default one drop.
Reno network with a default one drop.
Reno network with a default one drop.
SACK with a default one drop.
SACK with a default one drop.
28
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Overview
Overview
 The idea of TCP congestion control is for each resource to
determine how much capacity is available in the network, so
that it knows how many packets it can safely have in transit.
 TCP maintains a new state variable for each connection, called
CongestionWindowwhich is used by the source to limit how
much data it is allowed to have in transit at a given time. One
way to determine the CongestionWindow is to learn the level
congestion existed in the network.
 Early TCP implementations followed a go-back-n model using
cumulative positive acknowledgment and requiring a
retransmit timer expiration to resend data lost during
transportation. Modern TCP implementations contain a
number of algorithms aimed at controlling network congestion
while maintaining good user throughput.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Slow Start
Slow Start
vs
vs
Congestion Control
Congestion Control
 Slow-Start algorithm
 Add a congestion window, cwnd, to the per-connection state.
 When starting or restarting after a loss, set cwnd to one packet.
 On each ACK for new data, e.g. an ACK with a higher value then the
highest seen so far, increase cwnd by one packet.
 When sending, send the minimum of the receiver's advertised window
and cwnd.
 Congestion Avoidance Algorithm
 On any timeout, set cwnd to half the current window size.
 On each ACK for new data increase cwnd by 1/cwnd.
 When sending , send the minimum of the receiver's advertised window
and cwnd.
29
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Tahoe
Tahoe
-
-
Reno
Reno
 Tahoe TCP
Tahoe TCP implements Slow Start and Congestion Avoidance. It also uses
the Fast Retransmit algorithm that works the following way: After
receiving a small number of duplicate acknowledgments for the same TCP
segment, the data sender infers that a packet has been lost and retransmits
the packet without waiting for a retransmission timer to expire.
 Reno TCP
The Reno TCP is similar to the Tahoe TCP, except it also includes Fast
Recovery, where the current congestion window is "inflated" by the number
of duplicate ACK:s the TCP sender has received before receiving a new
ACK. In addition, the Reno TCP agent does not return to slow-start during
Fast Retransmit , instead it reduces the congestion window to half the
current window size.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Modeling and Simulation
Modeling and Simulation
 A simple network consists of one client and one server. The
client performs FTP application to download 1,600,000-byte
file from the server. You will observe how TCP performs
congestion avoidance. There are 4 simulation sequences:
 Slow Start – Congestion Avoidance without dropping
packet.
 Tahoe TCP implemented Fast Retransmit with one drop.
 Reno TCP implemented Fast Retransmit and Fast Recovery
with one drop.
 SACK TCP
30
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
1.Time occurred congestion:
At the time we force one packet dropped. The
network interprets the packet dropped as
congestion occurred.
2. Recovery Time:
Fast Retransmit and Fast Recover recover the lost
packet faster than the previous scenario in which
Fast Recover and Retransmit are disabled.
3. Congestion Window Size:
When Fast Retransmit and Recover are applied to
the network, congestion window size is set to half
of previous Congestion Window Size once
congestion occurred. This helps sending data
more quickly. If you collect the download time,
you would see that it takes 11.25s for client to
download 1.6MB data in default one drop and
10.05s in Fast Retransmit and Recover case.
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Simulation Results
Simulation Results
Flatten out.
When congestion occurs, Received Segment Ack Number will flatten out as
there is no new packet sent.
31
Phuong Phuong TuTu –– Spring ‘02Spring ‘02
Reference
Reference
 OPNET Tutorials
 OPNET Documentation
 OPNET workshop 2001
 Computer Network – L. Petersons
 CISCO – Online Tutorial