KINGDOM OF MOROCCO

Office for Vocational Training and Labor Promotion

Planning and optimization of the computer network

RESEARCH DIRECTION AND ENGINEERING DEFORMATION

ICT SECTOR
 KINGDOM OF MOROCCO

Office of Vocational Training and Labor Promotion

DIRECTION RESEARCH AND ENGINEERING DEFORMATION

ICT SECTOR
 Document title

Summary

1. Why analyze performance?........................................................2

1.1. Analysis of performance data and bottlenecks........2
2. Instructions to establish a baseline.................................................2
2.1. Factors to consider when determining a baselinee 3
2.1.1. Use of a baseline..........................................................3
3. Factors affecting network performance...........................................4
4. Determination of network performance...............................................5
4.1. Introduction to Ethernet Operation.................................6
4.2. Ethernet: throughput.....................................................6
4.3. Ethernet Utilization:.........................................................................7
5. Network performance assessment tools.........................................8
5.1. Traffic control tools and protocol analyzers............................8
6. Network upgrade considerations................................9
7. Ethernet User Instructions recommended.....................................10
8. Calculation of gross data rate..............................................................12
9. Planning for Future Growth.........................................................13
10. Network performance evaluation instructions.............................14

Document Vintage Page

OFPPT @ No translatable text provided.December 07 1 - 18
Document Title

1. Why analyze performance?

Performance analysis is a necessary part of maintenance.
preventive of your network. By regularly analyzing the performances
from your server over periods of time ranging from a few days to
a few weeks and a few months, you can draw a line of
based on network performance. This information constitutes
comparative data that will allow the analysis of the system and the
resolution of potential problems.

1.1. Analysis of performance data and

bottlenecks

The analysis of performance data is likely to reveal some

problems such as the creation of bottlenecks at
resources exposed to excessive demand. A bottleneck
it is noted when the performance of a resource harms the
system performance as a whole. The demand to which
this resource must face may exceed its capabilities to the point of
causing congestion of the four subsystems: memory, processor,
disk and network.
Here are some of the reasons why bottlenecks occur.
of strangulation:

The subsystems are insufficient, so it is necessary

to add or upgrade components. For example, a
Insufficient memory is a common cause of bottleneck.
of strangulation.
The sub-systems do not share the workloads.
also and must be balanced. For example, an old map
network installed on a new server can cause a
engorgement.
A subsystem is malfunctioning and needs to be replaced.
For example, an abnormally low throughput on a network card can
indicate a malfunction of this card.
A program monopolizes a particular resource. By
for example, it may be that a program written by a consultant does not
does not share memory correctly.
A subsystem is not configured correctly. For example,
an old multi-speed network card can be configured to
10 megabits per second (Mbits/s) instead of 100 Mbits/s.

Instructions for establishing a baseline

You establish a baseline from a dataset
collected over an extended period characterized by burdens of
work and variable but usual network connections. The line of

Document Vintage Page

OFPPT @ The provided text seems to be
December 07 and does 2
a file name not- contain
18 translatable content.
Document Title

base is a measure of the behavior of a specified resource in

normal activity period.

2.1. Factor to consider when

determination of a baseline

By determining a baseline, you must take into account the factors

following:

When you develop the baseline, it is important to

know the type of system load and its evolution in the
time. This allows you to associate types of loads with
the use of resources and to determine if the level of
performances during these periods are reasonable.

Collect this performance data over a period of time

extended use of low, medium, and maximum, in order to identify
acceptable system performances. The determination of these
Acceptable performances constitute your baseline.

By observing the evolution of performance over time, you

you may identify a performance model reflecting the impact
normal simultaneous logins or logouts
(at the beginning or at the end of the day) or the expected slowdown
during system backups. Based on performance
observed and an analysis of cause-and-effect relationships, you
can you determine the acceptable performances.

The baseline lines established at an early stage of the phase of

Deployment facilitates the resolution of deployment issues.

2.1.1. Use of a baseline

Using a baseline, you must take into account the factors

following:
To quickly identify and resolve bottlenecks
of the System, you have an interest in establishing a baseline as soon as possible
as early as possible.
Use the baseline to detect long-term changes.
term Intervening in usage models and making
necessary an increase in capacities.

Document Vintage Page

OFPPT @ 23125949.doc December 07 3 - 18
Document Title

3. Factors affecting performance of

network
Networks may have restrictions regarding the volume of
data they can transfer over a given period. When during the
network infrastructure planning, you must be aware of
following factors that may affect network performance.

The bandwidth is the maximum volume of data that can be transmitted.

to transit on a communication pathway during a period
data. It is often measured in bits per second (bps).

Latency is the time required for a packet to transit from the

source to destination. It is generally measured in seconds.

The throughput is the data transfer rate obtained by combining the

effects of bandwidth and latency. To simplify, let's say that the
Bandwidth is what you pay for and the speed is what you get.
actually dispose.

The capacity is the maximum data handling capability in

practice.

The line speed is the actual data transmission speed over

the cable once a transmission has begun. It is generally
measured in megabits per second (Mbps). The wired speed does not
does not represent the actual capacity or throughput that the local network can
take over (bits on the cable).

The usage is the percentage of duration occupied by the cable and

includes a successful and failed frame transmission (collisions)
and errors). The term cable can also be used. As a rule
general, inactive time + usage = 100 percent.

Instability corresponds to variable delays on a network.

The jabotage is a continuous flow of random data transmitted over a

network due to a malfunction.

A bottleneck is a delay that occurs when a part

a network is slower than the others and therefore hinders the overall bandwidth.

Collisions are frames whose transmission has failed on a medium.

shared because the senders attempted to send multiple frames
simultaneously.

Document Vintage Page

OFPPT @ No translatable text provided.December 07 4 - 18
Document Title

The yield is the percentage of raw data contained in

a frame in relation to the total frame length. Each frame
Ethernet includes data and header information, as well as a
gap between the frames. Depending on the size of the frame, the
percentage of overdraft represented by the header compared to the
raw data may vary. For important frames, compared to
weak frames, the overage accounts for a lower percentage of the
frames. Important frames are therefore more reliable than lesser frames.
small.

The frame rate represents the number of frames sent by

Second. The frame rate depends on the size of the frame.
The two following variables constitute the two main factors in
to take into account when calculating the throughput capacity on a network
Shared Ethernet:

Number of users: As the number of users

As the number increases, the number of collisions increases accordingly. A
maximum throughput is achieved when there are only two nodes, in the
in the context of a switched connection. The minimum rate is obtained
when the maximum number of users is 1,024.

Frame size: The rate decreases for small frames; it is

therefore less reliable. Important frames reduce the risks of
contention

4. Determination of network performance

Although an Ethernet network can operate at utilization rates

high, users will experience unacceptable delays as
that usage increases. Fixed delays, known as "
latency}, can be acceptable for certain types of applications

Document Vintage Page

OFPPT @ No text provided for translation.
December 07 5 - 18
Document Title

in which they must be less obvious. Variable delays,

known as 'instability', are unacceptable for
applications such as multimedia streaming.

Furthermore, the communication protocol stacks executing a

reliable delivery (TCP for example) await an acknowledgment from
frames to be returned during a given period. If the acknowledgment of
reception was not returned within the expected time frame, a
retransmission will take place. It is therefore important to determine the latency.
and the instability you wish to accept on your network.

4.1. Introduction Operation of Ethernet

On a shared standard Ethernet network, all nodes must

share the network support. The more the number of nodes connected to the network is
the more the network is busy; the availability of support decreases
and collisions are more frequent. Performance decreases at
as the throughput decreases and the latency increases. On a network
switched, the connection only involves two nodes. The collisions are
so rare and the flow can remain high because the support is not
shared by more than two nodes.

4.2. Ethernet: speed

The Ethernet throughput graph illustrates that the throughput decreases at
As the number of nodes increases, the small frames decrease.
also the flow.

A switched Ethernet (half-duplex) is always shared but, due to

that there are only two n.uds or connections, the collisions are much less
frequent. You can therefore use it reliably and permanently
at 80 percent.

The graph also indicates that, in an Ethernet environment

switched, the throughput can reach 90 percent of wired speed for
1,000 byte frames. A switched Ethernet environment uses a
switches to connect the network. A commuter is a bridge
intelligent and with multiple ports used to segment a local network
important in several smaller local networks.

The maximum flow depends essentially on the following two variables:

Number of users: The more users you have on a

local network, the more the channel reveals contention and the more the number
Collisions are important. The maximum throughput is reached with two
n.uds only on a local network, also called "
switched connection. This implies that both nodes can
generate the maximum frame rate.

Document Vintage Page

OFPPT @ 23125949.doc December 07 6 - 18
Document Title

Frame size: The Ethernet speed decreases in relation to

smaller frames. Conversely, important frames are larger.
effective and have the additional advantage of reducing the
risks of contention and collisions of the channel.

4.3. Ethernet Usage:

The Ethernet Usage graph (graph B) illustrates that latency

increases as usage increases.

Consider the following three usage trends:

You will need to quickly rectify the situation.

! Weak = 0-50 for cent of use
This level of usage reveals relatively few collisions (less than
of 10 percent) and low latency or instability.

! Moderate à high = 50-80 for cent of use

This level of usage is starting to produce significant delays. It
may be deemed acceptable for file transfer, but broadcasting
multimedia by stream produces an probably unacceptable instability.
This beach is acceptable for short periods, but you do not
you can't maintain this level permanently.

! Saturation = 80 percent or higher resolution This level

use leads to significant delays, sometimes exceeding
1 second. You may encounter situations in which
some nodes transmit significant streams of frames while
others wait several seconds. (This situation is called "
"congested or saturated network".) If you operate at this level of usage,
you must quickly rectify the situation.

Document Vintage Page

OFPPT @ December 07 7 - 18
Document Title

5. Performance assessment tools of

network
Most tools help you measure network load and
determine the use of the network. Some of these tools, like the tools
Software traffic control and protocol analyzers are free.
Others, like pocket diagnostic tools, are more user-friendly and
portable. Whatever tool is used, it provides you with information
useful on the network status.
Several relatively inexpensive software tools are now
available and offer tracking features and analyzers of
protocol.

5.1. Traffic control tools and analyzers

protocol

Control tools: Control tools allow you to measure

the use of the network, the distribution of protocol and the use of
bandwidth.

Protocol analyzers: You can use the analyzers of

protocol to study the content of network frames in order to proceed
to a complex network troubleshooting. Protocol analyzers can
also simulate traffic on the network to help you plan
future growth. The Network Monitor is a classic example
of this type of tool.

Network administration software: You can use software

network administration to perform protocol analysis and certain
types of controls. Other examples of this type of products are
notably Observer Suite from Network Instruments, LANdecoder32 from
Triticom and the Ethereal network analyzer, FlukNetwork

Hardware-assisted protocol analyzers

Hardware-assisted protocol analyzers are generally

tools requiring advanced knowledge and used to solve and
analyze networks. They are generally run on a computer
personnel and have a specific network card. Sniffer, from Network
Associates is the most well-known product in this range of tools.

Network management software

Most SNMP (Simple Network Management Protocol) network management platforms

Network Management Protocols) are available nowadays and most
Recent networking equipment allows for the collection of statistics.

Document Vintage Page

OFPPT @ No translatable text provided.December 07 8 - 18
Document Title

network. You will generally use the remote control function

(RMON) SNMP to collect information about the ports of
individual commuting. Here are some examples of platforms
network administration: CiscoWorks by Cisco Systems, Optivity by
Nortel Networks and OpenView from Hewlett-Packard.

Pocket diagnostic tools

Although pocket diagnostic tools generally come down to

pocket cable testers, these devices now integrate
Layer 2 and 3 network functions. Their portability and functions (testing of
wiring, protocol and frame analysis, control and diagnostics of
server, as well as general traffic measurements, for example) make these
devices of very useful tools, notably due to their ability to
measure network usage and collision rate at a specific location
of the network. Here are two examples of products of this type: FrameScope 350
from Agilent Technologies and Fluke's OneTouch Series II Network Assistant
Networks

6. Considerations for Upgrade

of the network

Introduction

You need to take into account many significant factors.

during the upgrade of your network. As an administrator
network, in addition to the future plans, you need to assess the use and the
current trends. You must then choose the level of
network performances that you wish to keep. Finally, you must
anticipate all possible repercussions that may occur
during the modification of the environment.

When defining the current usage statistics, make sure to

collect the average and maximum usage rates of the different segments
over a period of at least 24 hours. Knowing the situation
current state of your network gives you a foundation on which to base your
upgrade.

Current traffic sources

Once the use of network traffic is determined, it proves useful to

know the source and destination of the traffic. Ask yourself the questions
following:

Do hosts generate traffic according to a workstation configuration?

post?

Do they have access to a server?

Document Vintage Page

OFPPT @ No text provided for translation.
December 07 9 - 18
Document title

Make sure that all bottlenecks are not due to

the physical hardware infrastructure only. Problems can
are found on other sources, such as installation errors, the
non-compliance with specifications, inappropriate configuration, weak
node performances, an inappropriate positioning of the firewalls and
saturated file servers.

Future growth

It is preferable to include future network growth, such as the

demonstrate the following scenarios:
A service of your company anticipates significant growth.
soon; you may need to increase the
bandwidth on this segment.
A service from your company is working intensively on
multimedia files; you may need to keep some
average and maximum usage rates lower for this segment.

Determination of targets

Once your current situation and potential future are assessed, you need to
choose the target utilization rates. When defining these rates, you
can you determine if servers or users with power must
be isolated from other network hosts.

Remember that switched connections can work.

at usage rates much higher than connections
shared.

Repercussions

It is possible that, during an upgrade, you may change

the environment in such a way that other problems arise. You
you must think about the potential downstream repercussions before
upgrade. For example, the upgrade of a
Does the specific segment cause a bottleneck downstream?

7.Ethernet User Instructions

recommended

Document Vintage Page

OFPPT @ No translatable text provided.December 07 10 - 18
Document Title

When planning a network infrastructure, you must

determine the maximum volume of data that is transmitted during
peak periods when users access data and services.
Knowing this maximum traffic value allows you to choose
the appropriate network speed to maintain an acceptable throughput of
network.

Type of connection

The two basic categories of connection types are Ethernet

shared and switched Ethernet. Shared Ethernet is the conventional type.
in which all hosts are connected and are competing with each other
others for the bandwidth. On the contrary, a switched Ethernet
involves one or more point-to-point connections between hosts or the
segments. Devices connected to Ethernet using a
switches are not competing with each other and have
so a dedicated bandwidth. Switched Ethernet networks are
increasingly used because they are a reliable means
and practice of extending the bandwidth of existing Ethernet.

Wired speed

The wire speed represents the rate of data transfer that can be
reached at the level of physical wiring. The average usage is the
best rate, whether of average incoming traffic or average outgoing traffic,
divided by the transmission speed of a line.

Usage

The maximum utilization is the best rate, namely the maximum incoming traffic,
it is the maximum outgoing traffic, divided by the transmission speed of a
line.

Document Vintage Page

OFPPT @ 23125949.doc December 07 11 - 18
Document Title

Flow limits

The rate limits impose limits on the rate at which data

can be sent or received over a given period. The graph
proposed illustrates two different limitations of the flow: raw data and
maximum data.

Recommendations

Here are some recommendations that you can adopt in the context
of optimizing the bandwidth of your network:

Limit shared Ethernet networks to a usage of less than

50 percent and at a maximum usage of 80 percent.

A local network segment should not host more than 200.

users.

A maximum collision rate of 20 percent is acceptable.

A switched connection can be executed for use of

90 percent maximum because the connection must not
must necessarily be shared with other n.uds.

Full-duplex switched connections can be

executed at a maximum usage of 95 percent on each
pair of cables (transmission and reception) to achieve a use
theoretical of 190 percent.

8. Calculation of gross data throughput

Several methods are available to you to determine the gross flow of

data transfer. The data being contained in frames and
of packets containing header information, you must remove
the bits used for the overage if you measure the gross data rate.

Gross data rate

The gross data rate is the volume of usable data transferred.

to the advanced layer 3 software.

Gross data rate = NET USAGE (USAGE.

COLLISIONS) *efficiency*wire speed

Factors affecting theflow rate

Fast Ethernet has a rate of 100 Mbps. The data transfer rate
current user is well below; it can reach 50 percent of
specified data rate. Adding additional n.uds reduces the
recognized speed of a LAN Shared Ethernet because of collisions

Document Vintage Page

OFPPT @ The provided text appears toDecember
be a file name,
07 12 - 18
which cannot be translated. Please provide text for translation.
Document Title

producing, thus causing a delay in the nodes over a given period,

then an attempt at retransmission.
In addition, the header information of frames and packets reduced from
significantly the raw data rate. The headers contain
the source and destination addresses, information about
the establishment of a link, error control codes, etc. From
even some protocols require that the recipient of the frames and
packets, or groups of frames and packets, acknowledges it
reception. This creates additional traffic preventing sending
raw data.

Example

For example, a data packet can contain 53 bytes.

information but, since 5 bytes are reserved for
header information, only 48 bytes of raw data are actually
transmitted in the package. Generally, the more the information in-
the more numerous they are, the lower the volume of data sent.
The gross data rate is simply another characteristic.
network performances that you can determine through the
data on usage. It indicates the volume of useful data
actually transferred over the network.

9. Planning for future growth

Several useful instructions help you manage your growth.
current and to plan for future growth.

Planning for future growth

You must anticipate future growth and ensure that all

acquired products may be adapted to your long-term plans. By
for example, you would not want to replace a hardware component a
only after and encounter the same problems.

Use scalable products and technologies

Ensure that all products or network technologies can be set to

level and also be compatible upstream with your network
existing. Use products and technologies through which your
businesses can grow and are not at the end of their life.

Be proactive and not reactive

If you operate a reactive administration, some of your users

will be unhappy and, as a result, it will be too late to turn back
back. To be proactive, you need to regularly control the
usage and bandwidth data in order to take the measurements
appropriate before your users
do encounter a problem and before exceeding the bandwidth capacity
passerby.

Document Vintage Page

OFPPT @ The provided text is not translatable
December it appears to13
as 07 be -a 18
filename.
Document Title

Test the new applications

Be sure to test any new application to know its potential

impact on the network before its implementation. Make sure you know
the load of this application on your network before its implementation.

Stay informed about current technologies

Be informed about the network technologies you can use.

The products or technologies that may be available can
improve the reliability of your network and simplify its administration. From
You can also learn more through discussion groups or
publications that can help you improve your network.

10. Performance Evaluation Instructions

of the network

There are four simple instructions to follow when evaluating the flow of
network. These instructions provide you with the information you have
need to make appropriate decisions regarding the optimization of a
physical network.

Check the capabilities of current Ethernet technology

System engineers are strongly advised to keep themselves

informed about the advances in Ethernet technology. Through the assessment
existing technologies, you will be able to ensure that your
network benefits from the latest technological developments.

Compare these capabilities to those of the existing network

Once the current Ethernet technologies are analyzed, you must

compare the available elements of the elements implemented in
the infrastructure of your network. Assessing the current bandwidth allows you
to identify performance issues and offers you a foundation
correct on which you can estimate future growth.

Determine if current technologies can solve the

problems of the corporate network

The costs of implementing the latest technology are not worth it

sometimes not the performance improvements that you will benefit from.
You need to determine if the benefits justify the costs and if the implementation
In operation can solve the network performance issue
specific that you are encountering.

Determine the needs for future growth

Document Vintage Page

OFPPT @ 23125949.doc December 07 14 - 18
Document Title

Once the foundation is established, and when you know the future needs
of the company that will affect the network performance, you need to
estimate the future bandwidth needs of the network. Certain issues that could
Additional nodes that affect future growth, the addition
bandwidth-consuming applications and the requirements
additional access to resources.

Document Vintage Page

OFPPT @ No translatable text found. December 07 15 - 18
Document Title

To delve deeper into the subject....

Proposal of useful references to deepen the discussed theme

Reference sources
Cite the authors and reference sources used for the development of the
support

Document Vintage Page

OFPPT @ No text provided for translation.
December 07 16 - 18