LAN Design
LAN Design Goals and Components
LAN Design Goals
Designing a network can be a challenging task, and involves more than just connecting computers together. A
network requires many features in order to be scalable and manageable. To design reliable, scalable networks,
network designers must realize that each of the major components of a network has distinct design requirements.
Even a network that consists of only fifty nodes can pose complex problems that lead to unpredictable results.
Attempting to design and build networks that contain thousands of nodes can pose even more complex problems.
The first step in designing a LAN is to establish and document the goals of the design. These goals are particular to
each organization or situation. However, the following requirements tend to show up in most network designs:
Functionality-The network must work. That is, it must allow users to meet their job requirements. The network
must provide user-to-user and user-to-application connectivity with reasonable speed and reliability.
Scalability-The network must be able to grow. That is, the initial design should grow without any major
changes to the overall design.
Adaptability-The network must be designed with an eye toward future technologies, and it should include no
element that would limit implementation of new technologies as they become available.
Manageability-The network should be designed to facilitate network monitoring and management to ensure
ongoing stability of operation.
Critical Components of LAN Design
With the emergence of high-speed technologies such as Asynchronous Transfer Mode (ATM) and more complex
LAN architectures that use LAN switching and VLANs over the past several years, many organizations have been
upgrading existing LANs or planning, designing, and implementing new LANs. To design LANs for high-speed
technologies and multimedia-based applications, network designers should address the following critical
components of the overall LAN design:
The function and placement of servers
Collision detection
Segmentation
Bandwidth versus broadcast domains
The function and placement of servers when designing a network
One of the keys to designing a successful network is to understand the function and placement of servers needed for
the network. Servers provide file sharing, printing, communication, and application services, such as word
processing. Servers typically do not function as workstations; rather, they run specialized operating systems, such as
NetWare, Windows NT, UNIX, and Linux. Today, each server usually is dedicated to one function, such as e-mail
or file sharing.
Servers can be categorized into two distinct classes: enterprise servers and workgroup servers. An enterprise server
supports all the users on the network by offering services, such as e-mail or Domain Name System (DNS). E-mail or
DNS is a service that everyone in an organization (such as the Washington School District) would need because it is
a centralized function. On the other hand, a workgroup server supports a specific set of users, offering services such
as word processing and file sharing, which are services only a few groups of people would need.
Enterprise servers should be placed in the main distribution facility (MDF). This way, traffic to the enterprise
servers has to travel only to the MDF and does not need to be transmitted across other networks. Ideally, workgroup
servers should be placed in the intermediate distribution facilities (IDFs) closest to the users accessing the
applications on these servers. You merely need to directly connect servers to the MDF or IDF. By placing
workgroup servers close to the users, traffic only has to travel the network infrastructure to that IDF, and does not
affect other users on that network segment. Within the MDF and IDFs, the Layer 2 LAN switches should have 100
Mbps or more allocated for these servers.
Intranet
One common configuration of a LAN is an intranet. Intranet Web servers differ from public Web servers in that,
without the needed permissions and passwords, the public does not have access to an organization's intranet.
Intranets are designed to be accessed by users who have access privileges to an organization's internal LAN. Within
an intranet, Web servers are installed in the network, and browser technology is used as the common front-end to
access information, such as financial data or graphical and text-based data stored on those servers.
The addition of an intranet on a network is just one of many application and configuration features that can cause an
increase in needed network bandwidth over current levels. Because bandwidth has to be added to the network
backbone, network administrators should also consider acquiring robust desktops to get faster access into intranets.
New desktops and servers should be outfitted with 10/100-Mbps Ethernet network interface cards (NICs) to provide
the most configuration flexibility, thus enabling network administrators to dedicate bandwidth to individual end
stations as needed.
Why contention is an issue with Ethernet
You should decide carefully on the selection and placement of networking devices to be used in the LAN in order to
decrease the collision detection and media contention on a network. Contention refers to excessive collisions on
Ethernet caused by too many devices, each with a great demand for the network segment. The number of broadcasts
becomes excessive when there are too many client packets looking for services, too many server packets announcing
services, too many routing table updates, and too many other broadcasts dependent on the protocols, such as
Address Resolution Protocol (ARP).
An Ethernet node gets access to the wire by contending with other Ethernet nodes for the right to do so. When your
network grows to include more nodes on the shared segment or wire, and these nodes have more and more messages
to transmit, the chance that a node will contend successfully for its share of the wire gets much worse, and the
network bogs down. The fact that contention media access does not scale or allow for growth, is Ethernet's main
disadvantage.
As traffic increases on the shared media, the rate of collisions also increases. Although collisions are normal events
in Ethernet, an excessive number of collisions will (sometimes dramatically) reduce available bandwidth. In most
cases, the actual available bandwidth is reduced to a fraction (about 35% to 40%) of the full 10 Mbps. This
reduction in bandwidth can be remedied by segmenting the network by using bridges, switches, or routers.
How broadcast domains relate to segmentation
Segmentation is the process of splitting a single collision domain into two or more collision domains, as shown in.
Layer 2 (the data link layer) bridges or switches can be used to segment a logical bus topology and create separate
collision domains, which results in more bandwidth being available to individual stations. Notice in the figure that
the entire bus topology still represents a single broadcast domain because, although bridges and switches do not
forward collisions, they forward broadcast packets.
All broadcasts from any host in the same broadcast domain are visible to all other hosts in the same broadcast
domain. Broadcasts must be visible to all hosts in the broadcast domain in order to establish connectivity. The
scalability of the bandwidth domain depends on the total amount of traffic, and the scalability for a broadcast
domain depends on the total broadcast of the traffic. It is important to remember that bridges and switches forward
broadcast (FF-FF-FF-FF-FF) traffic, and that routers normally do not.
The difference between bandwidth and broadcast domains
A bandwidth domain is everything associated with one port on a bridge or switch. In the case of an Ethernet switch,
a bandwidth domain is also known as a collision domain. All workstations within one bandwidth domain compete
for the same LAN bandwidth resource. All the traffic from any host in the bandwidth domain is visible to all the
other hosts. In the case of an Ethernet collision domain, two stations can transmit at the same time, causing a
collision. (see picture next page)
Network Design Methodology
Gathering and analyzing requirements
For a LAN to be effective and serve the needs of its users, it should be designed and implemented according to a
planned series of systematic steps, which include the following:
Gathering the users' requirements and expectations
Analyzing requirements
Designing the Layer 1, 2, and 3 LAN structure (that is, topology)
Documenting the logical and physical network implementation
The first step in designing a network
should be to gather data about the
organizational structure. This information
includes the organization's history and
current status, projected growth, operating
policies and management procedures,
office systems and procedures, and the
viewpoints of the people who will be using
the LAN. You need to answer the
following questions: Who are the people
who will be using the network? What is
their level of skill, and what are their
attitudes toward computers and computer
applications? Answering these and similar
questions will help determine how much
training will be required and how many
people will be needed to support the LAN.
Ideally, the information gathering process helps clarify and identify the problems. You also need to determine
whether there are documented policies in place. Has some data been declared mission critical? Have some
operations been declared mission critical? (Mission-critical data and operations are those that are considered key to
businesses, and access to them is critical to the business running on a daily basis.) What protocols are allowed on the
network? Are only certain desktop hosts supported?
Next, you should determine who in the organization has authority over addressing, naming, topology design, and
configuration. Some companies have a central Management Information Systems (MIS) department that controls
everything. Some companies have very small MIS departments and, therefore, must delegate authority to
departments. Focus on identifying the resources and constraints of the organization. Organization resources that can
affect the implementation of a new LAN system fall into two general categories: computer hardware/software and
human resources. An organization's existing computer hardware and software must be documented, and projected
hardware and software needs identified. How are these resources currently linked and shared? What financial
resources does the organization have available? Documenting these types of things helps you estimate costs and
develop a budget for the LAN. You should make sure you understand performance issues of any existing network.
Factors that affect network availability
Availability measures the usefulness of the network. Many things affect availability, including the following:
Throughput
Response time
Access to resources
Every customer has a different definition of availability. For example, there may be a need to transport voice and
video over the network. However, these services require more bandwidth than is available on the network or
backbone. You can increase availability by adding more resources, but resources drive up cost. Network design
seeks to provide the greatest availability for the least cost.
After considering availability, the next step in
designing a network is to analyze the
requirements of the network and its users that
were gathered in the last step. Network user
needs constantly change. For example, as
more voice- and video-based network
applications become available, the pressure to
increase network bandwidth will become
intense.
Another component of the analysis phase is
assessing the user requirements. A LAN that is
incapable of supplying prompt and accurate
information to its users is of little use.
Therefore, you must take steps to ensure that
the information requirements of the
organization and its workers are met
Physical topologies used in networking
After determining the overall requirements for the network, the next step is to decide on an overall LAN topology
that will satisfy the user requirements. In this curriculum, we concentrate on the star topology and extended star
topology. As you have seen, the star/extended star topology uses Ethernet 802.3 carrier sense multiple access
collision detect (CSMA/CD) technology. The reason that this curriculum focuses on a CSMA/CD star topology is
that it is by far the dominant configuration in the industry.
The major pieces of a LAN topology design can be broken into three unique categories of the OSI reference model-
the network layer, the data link layer, and the physical layer. These components are discussed in the following
sections.
Layer 1 Design
Designing the layer 1 topology: signaling method, medium type, and maximum
length
In this section, you will examine Layer 1 star and extended star topologies.
The physical cabling is one of the most important components to consider when designing a network. Design issues
include the type of cabling to be used (typically copper or fiber) and the overall structure of the cabling. Layer 1
cabling media include types such as Category 5 unshielded twisted-pair (UTP) and fiber-optic cable, along with the
TIA/EIA-568-A standard for layout and connection of wiring schemes. In addition to distance limitations, you
should carefully evaluate the strengths and weaknesses of various topologies, as a network is only as effective as its
underlying cable. Most network problems are caused by Layer 1 issues. If you are planning any significant changes
for a network, you should do a complete cable audit to identify areas that require upgrades and rewiring.
Whether you are designing a new network or re-cabling an existing one, fiber-optic cable should be used in the
backbone and risers, with Category 5 UTP cable in the horizontal runs. The cable upgrade should take priority over
any other needed changes, and enterprises should ensure-without exception-that these systems conform to well-
defined industry standards, such as the TIA/EIA-568-A specifications.
The TIA/EIA-568-A standard specifies that every device connected to the network should be linked to a central
location with horizontal cabling. This is true if all the hosts that need to access the network are within the 100-meter
distance limitation for Category 5 UTP Ethernet, as specified by TIA/EIA-568-A standards. The table below lists
cable types and their characteristics.
Diagramming a standards-based Ethernet cable run from the workstation to the
HCC, including distances
In a simple star topology with only one wiring closet, the MDF includes one or more horizontal cross-connect
(HCC) patch panels. HCC patch cables are used to connect the Layer 1 horizontal cabling with the Layer 2 LAN
switch ports. The uplink port of the LAN switch, depending on the model, which is unlike other ports because it
does not cross over, is connected to the Ethernet port of the Layer 3 router using patch cable. At this point, the end
host has a complete physical connection to the router port.
HCC, VCC, MDF, IDF, and POP
When hosts in larger networks are outside the 100-
meter limitation for Category 5 UTP, it is not unusual
to have more than one wiring closet. By creating
multiple wiring closets, multiple catchment areas are
created. The secondary wiring closets are referred to
as IDFs. TIA/EIA 568-A Standards specify that IDFs
should be connected to the MDF by using vertical
cabling, also called backbone cabling. As shown in
figure , A vertical cross-connect (VCC) is used to
interconnect the various IDFs to the central MDF.
Because the vertical cable lengths typically are longer
than the 100-meter limit for Category 5 UTP cable,
fiber-optic cabling normally is used, as shown in
figure.
10BASE-T and 100BASE-TX Ethernet
Fast Ethernet is Ethernet that has been upgraded to 100 Mbps. This type uses the standard Ethernet broadcast-
oriented logical bus topology of 10BASE-T, along with the familiar CSMA/CD method for Media Access Control
(MAC). The Fast Ethernet standard is actually several different standards based on copper-pair wire (100BASE-TX)
and on fiber-optic cable (100BASE-FX), and it is used to connect the MDF to the IDF.
Elements of a logical topology diagram
As shown below, the logical diagram is the network topology model without all the detail of the exact installation
path of the cabling. It is the basic road map of the LAN.
Elements of the logical diagram include:
The exact locations of the MDF and
IDF wiring closets.
The type and quantity of cabling used to
interconnect the IDFs with the MDF,
along with how many spare cables are
available for increasing the bandwidth
between the wiring closets. For
example, if the vertical cabling between
IDF 1 and the MDF is running at 80%
utilization, you can use two additional
pairs to double the capacity
Detailed documentation of all cable
runs, as shown in Figure , the
identification numbers, and which port
on the HCC or VCC the run is
terminated on. For example, say Room
203 has lost connectivity to the network. By examining the cutsheet, you can see that Room 203 is running
off cable run 203-1, which is terminated on HCC 1 port 13. You can now test that run by using a cable
tester to determine whether the problem is a Layer 1 failure. If it is, you can simply use one of the other two
runs to get the connectivity back and then troubleshoot run 203-1.
Layer 2 Design
Common Layer 2 devices and their impact on network domains
The purpose of Layer 2 devices in the network is to
provide flow control, error detection, error
correction, and to reduce congestion in the network.
The two most common Layer 2 devices (other than
the NIC, which every host on the network must
have) are bridges and LAN switches. Devices at this
layer determine the size of the collision domains and
broadcast domains. This section concentrates on the
implementation of LAN switching at Layer 2.
Asymmetric switching
Collisions and collision domain size are two factors that
negatively affect the performance of a network. By using LAN
switching, you can microsegment the network, thus eliminating
collisions and reducing the size of collision domains. Another
important characteristic of a LAN switch is how it can allocate
bandwidth on a per-port basis, thus allowing more bandwidth to
vertical cabling, uplinks, and servers. This type of switching is
referred to as asymmetric switching, and it provides switched
connections between ports of unlike bandwidth, such as a
combination of 10-Mbps and 100-Mbps ports.
The effect microsegmentation can have on a network
Microsegmentation means using bridges and switches to boost performance for
a workgroup or a backbone. Typically, boosting performance in this manner
involves Ethernet switching. Switches can be used with hubs to provide the
appropriate level of performance for different users and servers.
Determining the number of cable runs and drops
By installing LAN switching at the MDF and IDFs and vertical cable between the MDF and the IDFs, the vertical
cable is carrying all the data traffic between the MDF and the IDFs; therefore, the capacity of this run must be larger
than that of the runs between the IDFs and workstations. Horizontal cable runs use Category 5 UTP, and no cable
drop should be longer than 100 meters, which allows links at 10 Mbps or 100 Mbps. In a normal environment, 10
Mbps is adequate for the horizontal cable drop.
Because asymmetric LAN switches allow for mixing 10-Mbps and 100-Mbps ports on a single switch, the next task
is to determine the number of 10-Mbps and 100- Mbps ports needed in the MDF and every IDF. This can be
determined by going back to the user requirements for the number of horizontal cable drops per room and the
number of drops total in any catchment area, along with the number of vertical cable runs. For example, say user
requirements dictate that 4 horizontal cable runs be installed to each room. The IDF that services a catchment area
covers 18 rooms. Therefore, 4 drops ×18 rooms = 72 LAN switch ports.
Determining the size of collision domains in hubbed and switched networks
To determine the size of a collision domain, you must determine how many hosts are physically connected to any
single port on the switch. This also affects how much network bandwidth is available to any host. In an ideal
situation, there is only one host connected on a LAN switch port. This would make the size of the collision domain 2
(the source host and destination host). Because of this small collision domain, there should be almost no collisions
when any two hosts are communicating with each other. Another way to implement LAN switching is to install
shared LAN hubs on the switch ports and connect multiple hosts to a single switch port. All hosts connected to the
shared LAN hub share the same collision domain and bandwidth.
Note that some older switches (e.g. Cisco’s Catalyst 1700) don't truly support sharing the same collision domain and
bandwidth because they don't maintain multiple MAC addresses mapped to each port. In that case, there are many
broadcasts and ARP requests.
Diagramming hub placement in a standards-based extended star topology
Shared-media hubs are generally used in a LAN switch environment to create more connection points at the end of
the horizontal cable runs. This is an acceptable solution, but you must ensure that collision domains are kept small
and bandwidth requirements to the host are accomplished according to specifications gathered in the requirements
phase of the network design process.
Migrating a network from 10 Mbps
to 100 Mbps
As the network grows, the need for more
bandwidth increases. In the vertical cabling
between MDF and IDFs, unused fiber optics can
be connected from the VCC to 100 Mbps ports
on the switch. The network shown doubles the
capacity of the vertical cabling in the network in
the following graphic by bringing up another
link.
In the horizontal cabling, you can increase the
bandwidth by a factor of 10 by repatching from
the HCC to a 100 Mbps port on the switch and
changing from a 10 Mbps hub to a 100 Mbps
hub. When sizing the Layer 2 LAN switch, it is
important to make sure there are enough 100
Mbps ports to allow for this migration to higher
bandwidth. It is important to document the speed
at which each active cable drop is running.
Layer 3 Design
Using routers as the basis for layer 3 network design
As shown in the Figure below, Layer 3 (the network layer) devices, such as routers, can be used to create unique
LAN segments and allow communication between segments based on Layer 3 addressing, such as IP addressing.
Implementation of Layer 3 devices, such as routers, allows for segmentation of the LAN into unique physical and
logical networks. Routers also allow for connectivity to wide-area networks (WANs), such as the Internet.
Layer 3 routing determines traffic flow between unique physical network segments based on Layer 3 addressing,
such as IP network and subnet. The router is one of the most powerful devices in the network topology.
As you have learned, a router forwards data packets based on destination addresses. A router does not forward LAN-
based broadcasts such as ARP requests. Therefore, the router interface is considered the entry and exit point of a
broadcast domain and stops broadcasts from reaching other LAN segments.
How VLANs can create smaller broadcast domains
One important issue in a network is the total number
of broadcasts, such as ARP requests. By using
VLANs, you can limit broadcast traffic to within a
VLAN and thus create smaller broadcast domains.
VLANs can also be used to provide security by
creating the VLAN groups according to function.
As shown above, a physical port association is used to
implement VLAN assignment. Ports P0, P1, and P4
have been assigned to VLAN 1. VLAN 2 has ports
P2, P3, and P5. Communication between VLAN 1 and
VLAN 2 can occur only through the router. This
limits the size of the broadcast domains and uses the
router to determine whether VLAN 1 can talk to
VLAN 2. This means you can create a security scheme based on VLAN assignment.
How a router provides structure to a network
Routers provide scalability because they can serve as firewalls for broadcasts. In addition, because Layer 3
addresses typically have structure, routers can provide greater scalability by dividing networks and subnets,
therefore, adding structure to Layer 3 addresses. The ways in which greater scalability in networks can occur are
shown in the table.
When the networks are divided into subnets, the final step is to develop and document the IP addressing scheme to
be used in the network. Routing technology filters data-link broadcasts and multicasts. By adding router ports with
additional subnet or network addresses, you can segment the internetwork as required. Network protocol addressing
and routing provide built-in scaling. When deciding whether to use routers or switches, remember to ask, "What
problem am I trying to solve?" If your problem is protocol related rather than contention oriented, then routers are
appropriate. Routers solve problems with excessive broadcasts, protocols that do not scale well, security issues, and
network-layer addressing. Routers, however, are more expensive and harder to configure than switches.
Why large, scalable LANs need to incorporate routers
Routers can be used to provide IP subnets to add structure to addresses. With bridges and switches, all unknown
addresses must be flooded out of every port. With routers, hosts using protocols with network-layer addressing can
solve the problem of finding other hosts without flooding. If the destination address is local, the sending host can
encapsulate the packet in a data-link header and send a unicast frame directly to the station. The router does not see
the frame and, of course, does not need to flood the frame. The sending host might have to use ARP. This would
cause a broadcast, but the broadcast is only a local broadcast and is not forwarded by the router. If the destination is
not local, then the sending station transmits the packet to the router. The router sends the frame to the destination or
to the next hop, based on its routing table. Given this routing functionality, it is clear that large, scalable LANs need
to incorporate some routers.
Diagramming a standards-based LAN that uses routers
The Figure shows an example of an implementation that has multiple physical networks. All data traffic from
Network 1 destined for Network 2 has to go through the router. In this implementation, there are two broadcast
domains. The two networks have unique Layer 3 IP addressing network/subnetwork addressing schemes. In a
structured Layer 1 wiring scheme, multiple physical networks are easy to create simply by patching the horizontal
cabling and vertical cabling into the appropriate Layer 2 switch using patch cables. This implementation provides
for robust security implementation. In addition, the router is the central point in the LAN for traffic destination.
Logical and physical network maps
After you have developed the IP addressing scheme for the customer, you should document it by site and by
network within the site. A standard convention should be set for addressing important hosts on the network. This
addressing scheme should be kept consistent throughout the entire network. By creating addressing maps, you can
get a snapshot of the network. Creating physical maps of the network helps you troubleshoot the network.
Logical Network Maps