Network

Architecture
Wednesday, September 27, 2006

Page 1
 What is Network Architecture?
 A network architecture is essentially a blueprint for how the
hardware and software in a network should be deployed
 LAN architecture is the overall design of a LAN. It includes:
• LAN hardware
• LAN software
• LAN topology
• Media access control (MAC) protocol
 The LAN’s network operating system is sometimes also
considered to be part of LAN architecture
 WAN architectures subsume WAN topologies, WAN services, and
the software and protocols used to facilitate resource sharing
among hosts and communication between nodes and terminals
 WAN architectures also include WAN switching and transmission
architectures such as circuit switching and packet switching
 Major WAN architectures include:
• Centralized network architectures
• Distributed processing systems
• Client/server architectures

Page 2
 Distributed systems
 File Transfer
 Peer – to – peer
 Client Server

Page 3
 Client/Server Computing Systems

 Client/server (C/S) computing is an application processing

framework in which the data processing load is divided
among several processes called clients and servers. Clients
issue requests to servers which provide specialized services
such as database processing
• Both client and server processes can exist in the same computer
 When clients and servers are located in different computers,
application processing is distributed over multiple
computers making the network the computer
• C/S configurations are common in LANs

Page 4
LAN C/S Computing Environment

Page 5
 International Networks
 Many WAN architectures are international in scope
 Some are used to support virtual teams; others are used to
overcome time-zone differences, facilitate international funds
transfers, and to support international supply chains
 Organizations face many challenges in developing and managing
international networks including:
• Politics
• Security
• Regulations
• Hardware
• Language
• Problem troubleshooting and resolution
• Coordination and control

Page 6
 The Internet’s Impact on WAN
Architectures
 The Internet is having a significant impact on WAN
architectures. For example, it has:
• Facilitated the emergence of 3- and n-tiered C/S architectures
• Led to increased server-based processing via thin clients that is
reminiscent of centralized, host-based processing
• Become an important aspect, if not the centerpiece, of the overall
network architecture of a wide variety of firms (especially for
their international network infrastructures)

Page 7
 LAN Basics

 LAN Standards provide basis for compatibility

across vendors and product lines
 Purpose of LAN - move data back and forth
between users, workstations, & servers
 IEEE - Institute of Electrical and Electronics
Engineers -
• Their Slogan: Networking the World
• The IEEE promotes the engineering process of creating,
developing, integrating, sharing, and applying knowledge
about electro- and information technologies and sciences for
the benefit of humanity and the profession.

Page 8
 LAN Basics

 OSI - Seven Layer Reference Model

• Layers are self-contained and isolated from each other
• Changes in one should not affect others
 IEEE 802 defines LAN protocols covering the
lowest two layers (physical and data link)
 Data Link divided into sublayers
• Logical Link Control (LLC)
• Media Access Control (MAC)
 Also Network Management functions

Page 9
 Protocol Architecture
 Lower layers of OSI model
 IEEE 802 reference model
 Physical
 Logical link control (LLC)
 Media access control (MAC)

Page 10
IEEE 802 v OSI

Page 11
 802 Layers -
Physical
 Encoding/decoding
 Preamble generation/removal
 Bit transmission/reception
 Transmission medium and topology

Page 12
 802 Layers -
Logical Link Control
 Interface to higher levels
 Flow and error control

Page 13
 802 Layers -
Media Access Control
 Assembly of data into frame with address and
error detection fields
 Disassembly of frame
• Address recognition
• Error detection
 Govern access to transmission medium
• Not found in traditional layer 2 data link control
 For the same LLC, several MAC options may
be available

Page 14
LAN Protocols in Context

Page 15
 802.3 MAC and Physical
 Medium Access Control - required to allow
stations to share transmission facility
 Interfaces with LLC sublayer and the physical
layer below
 IEEE 802.3 - Carrier-Sense Multiple Access with
Collision Detection (CSMA/CD)

Page 16
 IEEE 802.5 - Token Ring

 IBM
 Stations connected serially, bit-by-bit transfer from one
node to another
 MAC
• Capturing passing token
• Add data to token then transmit to next node
• Node by node transmission until it reaches destination node
• Destination nodes transmits original token and data. Sending station
is responsible for removing data and issuing new free token

Page 17
 IEEE 802.5 - Token Ring

 Monitor Station - monitors tokens and frames to make sure

rules are followed.
 Frame Types - different frame types for control or
information exchange

Page 18
 IEEE 802.5 - Token Ring
 T bit - token or frame
 Frame Status - contains bits to indicate status of frame.
• A bit = address recognized
• C bit = frame copied
 32 bit CRC
 Token Holding Timer controls/limits transmission by a
single station

Page 19
 IEEE 802.5 - Token Ring

 MAC Functions
• Frame Transmission -
– When MAC layer receives data from higher layer it adds header
and places in transmit queue
– Gets next free token , attaches data and transmits
– After station receives acknowledgement of successful receipt of
its last frame it places a free token on line
• Stripping - stations must remove its transmitted frames from ring
• Frame Reception - all stations read and repeat incoming frames. Also
monitor for control frames and frames with address that match its own.

• Beaconing and Neighbor Notification

Page 20
 IEEE 802.5 - Token Ring
 MAC (cont)
• Beaconing and Neighbor Notification - mechanism to allow all stations
to know address of upstream station
• Token Maintenance - Monitor Station can detect uncontrolled frames by
setting M bit to 1 as it passes. If it sees a frame with M bit set it knows
it wasn’t removed by sending station.
 Concentrators (MAU)- are interconnected by trunk cable

Page 21
 LAN Topologies
 There are two types of LAN topologies: physical and logical
 Physical LAN topology refers to the physical layout of the
network
• The way in which the communication is configured and how nodes
attach to the network
• Because the focus is on physical connections among hardware
component, physical topologies correspond to the physical layer of the
OSI reference model
 Logical topology is concerned with how messages are
passed from node to node within the network

Page 22
 Physical Topologies
 LAN’s have three basic physical topologies:
• Bus: all nodes attach to a common communication pathway or channel
• Ring: the medium forms a loop to which all nodes are attached
• Star: uses a central station (hub or switch) to which all other nodes have
point-to-point connections; all communication among nodes occurs
through this central station
 These are illustrated in Figure 8-1
 Physical star topologies are most common in today’s LANs

Page 23
Page 24
 Bus Topologies
 In a classic bus topology, the medium consists of a single wire or
cable to which other nodes are attached via connectors or
transceivers
• Variations include a primary bus with spurs (see Figure 8-2)
• Disadvantages include the potential for loose connections or breaks in the bus to
disrupt the entire network
 Early Ethernet LAN implementations were typically physical bus
architectures; today, most Ethernet implementations are physical
stars
• However, an Ethernet shared media hub is sometimes called a “bus in a box”
 Both IEEE 802.3 standard and IEEE 802.4 standards and their
protocols address communication over LANs with bus topologies

Page 25
 Ring Topologies
 In a physical ring topology, the communication medium
forms a closed loop (ring) and all stations are connected to
the loop
• Data is transmitted node-to-node in one direction on the ring (see Figure
8-3)
• Similar to a linear bus, the entire network could be disrupted if one of the
connectors or links in the ring should fail
 Physical ring topologies are less common than bus or star
topologies
 Token ring and FDDI LANs have physical ring topologies

Page 26
Page 27
 IEEE 802.5 and 802.6 LANs
 The most widely used microcomputer ring network is the
token passing ring. It conforms to the IEEE 802.5 standard
 Token ring networks physically look like a star topology, but
technically they are physical rings
• Token ring nodes attach to multistation access units (MAUs) – see Figure
8-4
• MAUs can be described as “a ring in a box”, because nodes attach to the
physical ring by connecting to the MAU (see Figure 8-5)
• MAUs can be interconnected to form larger rings (see Figure 8-6)
 IEEE 802.6 addresses dual-ring metropolitan area network
(MAN) architectures (see Figures 8-24 and 8-25)

Page 28
Figure 8-4

Page 29
Figure 8-5

Page 30
Figure 8-6

Page 31
 Star Topologies
 In LANs with star topologies, all nodes are connected to some kind
of wiring center such as a hub or switch (see Figure 8-7)
• Today, most LAN implementations physically resemble star topologies
 Each node is isolated on its own network segment in a physical star
topology which minimizes the possibility of total network disruption
by a malfunctioning connector, NIC, or link
• However, the network is vulnerable to wiring center failure
 The use of central connection points also facilitates network traffic
monitoring and network management, including network security
management
 ARCnet was one of the first LAN architectures with a star topology
(see Figure 8-8b)

Page 32
Figure 8-7

Page 33
 Logical Topologies
 Every LAN has both a physical and logical topology
 A LAN’s logical topology specifies how messages are passed from
node to node within the network
• It corresponds to the media access control (MAC) protocol used in the LAN
 Two logical LAN topologies exist:
• Sequential (or logical ring): data is passed from one node to another in a ring-like
sequence
– Token passing in token ring and FDDI LANs are examples
• Broadcast: nodes transmit frames/packets to all other nodes in the network; only
the intended recipient processes the entire frame/packet
– Ethernet LANs use a logical broadcast topology

Page 34
 Data Link Protocols
 Data link protocols, including those used in LANs, are responsible
for establishing the rules by which nodes gain access to a
network’s communication medium and exchange messages. Such
protocols describe several important aspects of the message
exchange process including:
• Delineation of data
• Error control
• Addressing
• Transparency
• Code independence
• Media access--this is governed by media access control (MAC) protocols

Page 35
 LLC and MAC Sublayers
 LANs employ two primary data link protocols:
contention and token passing
 In IEEE 802 standards, the data link layer is divided
into two sublayers LLC and MAC (see Figure 8-10)
• LLC (logical link control) is responsible for flow control,
message sequencing, message acknowledgement, and error
checking
• MAC (media access control) enables network nodes to access
the communication medium via contention or token passing

Page 36
Figure 8-10

Page 37
 CSMA/CD
 CSMA/CD (Carrier Sense Multiple Access and Collision Detection) is
most widely used contention-based MAC used in LANs
• It is the MAC protocol used in Ethernet LANs
 In a true contention MAC (like CSMA/CD), each node has equal
access to the medium
 As noted in Table 8-1, each node monitors the medium for data traffic
and if none is detected, it begins transmitting data
 A collision occurs when two or more nodes begin to transmit at the
same time
 To avoid collision recurrence, each node waits a random time
interval (hardwired in its NIC) before attempting to retransmit

Page 38
Table 8-1

Page 39
 CSMA/CA
 CSMA/CA (Carrier Sense Multiple Access and Collision
Avoidance) is a variation of CSMA/CD used in wireless
LANs because it is difficult to detect collisions in such
networks
 When CSMA/CA is used, each node must wait a
random time interval (hardwired in the wireless NIC)
after detecting a clear medium before transmitting

Page 40
 Token Passing
 Token passing is the other major MAC protocol found in LANs
 It is used in token ring and FDDI LANs and other networks with
logical ring topologies
 The token is a pre-defined bit pattern that is passed among network
attached computers until one of them wants to use the medium to
transmit data
 Token passing is summarized in Table 8-2
 In token ring networks that resemble physical star topologies, token
passing takes place within MAUs (see Figure 8-12)
 Token passing can be used in bus topologies as well as in physical
ring topologies (see Figure 8-11)
 Table 8-3 compares token passing and CSMA/CD

Page 41
Table 8-2

Page 42
Figure 8-12

Page 43
Table 8-3

Page 44
 Physical Layer Data Encoding
 Baseband transmission is common in LANs
 When LAN nodes use the communication medium, their NICs
transmit digital signals to represent the bits in data link layer
protocols frames directly onto the medium
 Some of the encoding schemes used in LANs are illustrated
in Figure 8-13. These include:
• Manchester encoding (used in 10 mbps Ethernet)
• 4B5B (used in 100 mbps Ethernet)
• Differential Manchester encoding (used in token ring LANs)
• NRZI (used in FDDI LANs)

Page 45
Figure 8-13

Page 46
 Ethernet LAN Architectures
 IEEE 802.3-compliant LANs are better known as Ethernet
LANs
 There are a variety of IEEE 802.3-compliant LANs (see Table
8-4)
 Today, most Ethernet LANs have physical star topologies;
some have physical bus topologies
 All Ethernet LANs have broadcast logical topologies and use
CSMA/CD as the MAC protocol
 Figure 8-9 illustrates widely used Ethernet frame formats

Page 47
Table 8-4

Page 48
Figure 8-9

Page 49
 Some Key Ethernet
Implementations
 Particularly important Ethernet implementations
include:
• Fast Ethernet (e.g. 100BaseT and 100BaseFX)
– The IEEE 802.3u specification covers Fast Ethernet
• Gigabit Ethernet (e.g. 1000BaseT, 1000BaseSX, and 1000BaseLX)
– The IEEE 802.3z specification addresses Gigabit Ethernet
• Iso-Ethernet enables Ethernet LANs at different geographic
locations to be connected over ISDN (see Figure 8-14)
– IEEE 802.9a addresses Isochronous Ethernet

Page 50
Figure 8-14

Page 51
 Token Ring Architectures
 Token ring networks are addressed in the IEEE 802.5
specification
 Physically, token ring LANs resemble star topologies,
but technically they are rings
• Ring is physically implemented in MAUs
• UTP is the most common cabling
 Speed is typically 16 mbps, however, 4 mbps and 100
mbps token ring networks exist
 IEEE 802.5 frame formats are illustrated in Figure 8-15

Page 52
Figure 8-15

Page 53
 FDDI LANs
 Fiber Distributed Data Interface (FDDI) was first recognized in
ANSI’s X3T9.5 specification
 Physically, it has a dual ring topology
 It has a sequential/ring logical topology and uses a variation
of token passing as the MAC protocol
 Key FDDI technologies are identified in Figure 8-16
• These include single attached stations (SAS), dual attached stations
(DAS), FDDI concentrators, and FDDI/Ethernet bridges
 FDDI is often used as a backbone network architecture (see
Figure 8-18)

Page 54
Figure 8-16

Page 55
 100VG-AnyLAN
 100VG-AnyLAN (aka 100BaseVG) is capable of transporting
both IEEE 802.3 and IEEE 802.5 frames
 It provides a mechanism for interconnecting 100 mbps
token ring and 100BaseT Ethernet LANs via specialized
hubs and routers (see Figure 8-20)
• 100VG-AnyLAN-compliant adapters are also needed
 It uses demand priority access (DPA) rather than CSMA/CD
as the MAC protocol in order to enable real-time voice and
video frames to be given priority over other data frames

Page 56
Figure 8-20

Page 57
 ATM LANs
 ATM (asynchronous transfer mode) is a switched network
architecture that employs 53-octet cells to transmit data
 Two data link layers are defined:
• ATM adaptation layer (AAL)
• ATM
 ATM physical topologies are stars
 ATM NICs with 25 speeds of 25, 100, or 155 mbps are
available for workstations
 Ethernet and token ring LANs can interface with an
organization’s ATM backbone network via ATM
gateway/access switches (see Figure 8-21)

Page 58
Figure 8-21

Page 59
 Wireless LAN Architectures
 IEEE 802.11x standards are the most important wireless LAN (WLAN)
specifications that exist today (see Table 8-6)
 WLANs are typically implemented as physical stars
• Nodes connect to wireless hubs called access points
 CSMA/CA is the MAC protocol for IEEE 802.11-compliant LANs
 IEEE 802.11 addresses FHSS (frequency hopping spread spectrum),
DSSS (direct sequence spread spectrum), and diffuse infrared
transmission
• User “roaming” capabilities are also addressed
 WiFi (Wireless Fidelity) certification has been developed to promote
interoperability among WLAN products
 The WISPR (Wireless ISP Roaming) standard is designed to enable
users to roam from one publicly accessible WLAN to another

Page 60
Table 8-6

Page 61
 Choosing Among LAN
Architectures
 A number of factors should be considered when selecting among LAN
architectures
 Some of the major factors are described in Table 8-7; others are described in
Table 8-10)
 Especially important factors to consider include:
• Immediate and recurring LAN costs (see Table 8-8)
• Total cost of ownership (TCO)
• Number of concurrent users that can be supported
• Transmission speed and data throughput
• Vendor support
• Manageability
• Scalability/expandability
• Security
• Adherence to widely accepted standards

Page 62
Table 8-7

Page 63
IEEE LAN Standards & Committees
 802.1 High-Level Interface
 802.2 Logical Link Control
 802.3 CSMA/CD
 802.4 Token Bus
 802.5 Token Ring
 802.6 Metropolitan Area
Networks (MANs)
 802.7 Broadband Technical
Advisory Group
 802.8 Fiber Optic Technical
Advisory Group
 802.9 Integrated Data and
Voice Networks
 802.10 LAN Security
 802.11 Wireless LANs
 802.12 Demand Priority
Access Method
Page 64
LAN Design

Page 65
 Introduction
 One of the most critical steps to insure a fast and stable
network is the design of the network. This design activity is
truly an in-depth process, which includes:
 Gathering the users requirements and expectations
 Determining data traffic patterns now and in the future based on growth
and Server placements
 Defining all of the layer 1, 2 &3 devices and along with LAN and WAN
topology
 Document the physical and logical network implementation

Page 66
 LAN Design Goals
 Functionality - the network must work with reasonable speed
and reliability.
 Scalability - the network must be able to grow without any
major changes to the overall design.
 Adaptability - the network must be designed with an eye
toward future technologies, and should include no element
that would limit implementation of new technologies as they
become available.
 Manageability - the network would be designed to facilitate
network monitoring and management.

Page 67
 Design Methodology
 Analyse requirements
 Develop LAN structure (topology)
 Set up addressing (and naming conventions) and routing

Page 68
 Analyse Requirements
 Business issues
 Technology issues
 Administrative issues
 Gather Data -
 Corporate Structure
 Business information flow
 Applications in use
 Current topology

Page 69
 Analyse Requirements
 Performance characteristics of current network
 Determine if documented policies are in place
 Mission-critical data
 Mission-critical operations
 Approved protocols and platforms
 Control versus distributed authority

Page 70
 Analyse Requirements
 Business requirements
 Technical requirements
 New applications or business operations
 Availability requirements
 Throughput
 Response time
 Access to resources

Page 71
 Design Rules
 Design Rule:
• “First and foremost, you must understand your
customer”
 Design Rule:
• “Find out what availability means to your customer”

Page 72
 Step 2–Develop LAN Topology
 LAN topology that will satisfy Step 1
requirements
 Star Topology
 Extended Star Topology
 Ring Topology
 Bus Topology
 Mesh Topology

Page 73
 Topology
 Bus Topology
• Rarely used in the design of new networks.
• Typically uses coaxial cable to connect hosts in a bus formation
• Difficult in some circumstances to add new hosts
 Star Topology
• Uses twisted pair cable to connect the NIC of all hosts to a central point,
such as a hub or switch
• Main method of building LAN’s today
• Easy to design and implement
• Addition of new hosts is easier than with bus topology

Page 74
 Topologies
 Extended Star Topology
• Uses a hierarchical structure
• Multiple smaller star networks are connected to form a larger LAN
• Typically, for example, each room in a building may form one star
network. The rooms are then connected to each other via a switch, to
form an extended star network.

Page 75
Page 76
Page 77
Page 78
 Step 3 – Layer 3 Addressing
 The router divides subnets and networks
 The router structures an internetwork
 Logical addressing should be mapped to the physical
network
 Develop and document the IP addressing scheme to be used
in the network
 Use consistent naming and addressing conventions to avoid
confusion

Page 79
 Setting up a VLAN
 Group users by department, team, or application
 Provide broadcast containment and security
 Routers provide communication between VLANs
 Enhanced security comes from the fact that communication
between VLAN’s is not possible except through a router

Page 80
 What are the problems?
 Media contention
 too many devices, all with a high demand for the network segment
 Excessive broadcasts
 client packets looking for services
 too many server packets announcing services
 too many routing table updates
 Need to transport new payloads
 voice and video network services

Page 81
 What are the problems?
 Need for more bandwidth
 Overloaded backbone
 Network-layer addressing issues
 running out of IP addresses
 need for physically separate subnets
 other issues dependent on the protocols

Page 82
 Review of Ethernet
 Segmentation - Bridging and switching are both used for
segmentation
 Results in multiple collision domains
 Still a single broadcast domain
• Stations can get dedicated bandwidth

 Collision domain (bandwidth domain) - In Ethernet, the network area

within which frames that have collided are propagated (everything
associated with one port on a bridge or switch). Repeaters and hubs
propagate collisions; LAN switches, bridges and routers do not .

Page 83
 More Ethernet
 Broadcast domain - The set of all devices that will receive broadcast
frames originating from any device within the set. Broadcast
domains are typically bounded by routers because routers do not
forward broadcast frames (everything associated with one port on a
router).

 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.

Page 84
 Network Design
 Layer 1 - Physical Layer
• Includes wire media type such as CAT5 UTP and fiber-optic cable along
with EIA/TIA 568 Standard for layout and connection of wiring schemes.

 Design Goal
• “Build this layer of the OSI model with speed and expansion capabilities”

Page 85
 Network Design
 Layer 2 - Data Link Layer
• Includes selection of Layer 2 devices such as bridges or LAN switches
used to interconnect the Layer 1 media to for a LAN segment. Devices at
this layer will determine the size of the collision and broadcast domains.
 Design Goals
• Create a concentration point within the MDFs or IDFs where end host can
be grouped at Layer 1 to form a physical LAN segment.
• Install LAN switching devices that use microsegmentation in order to
reduce the collision domain size.

Page 86
 Network Design
 Layer 3 – Network Layer
• Includes selection of layer 3 devices such as routers which are used to
create unique LAN segments and allow communication between
segments based on layer 3 addressing such as IP addressing.

 Design Goals
• Build a path between LAN segments that will filter the flow of data
packets.

Page 87
 Wiring & Topology
 Copper Wiring
• Coaxial Cable
• Twisted Pair Cable
 Fiber
• Multimode / Monomode (Single Mode)
• Typically used on large network backbones
 Wiring configurations
• Star
• Extended Star

Page 88
 Star Topology
 Star Topology using CAT6 (UTP):
 Specified by EIA/TIA 568 standard
 The MDF is the center of the star
 The MDF is the Point of Presence (POP) for outside services from the
WAN provider

Page 89
 Some terminology
 Vertical cabling - (Backbone cabling) Cabling that provides
interconnections between wiring closets, wiring closets and the POP,
and between buildings that are part of the same LAN.
 Catchment areas - Zone that falls within area that can be served by
an internetworking device such as a hub.
 MDF - Main Distribution Facility. Primary communications room for a
building. Central point of a star networking topology where patch
panels, hub, and router are located.
 IDF - Intermediate Distribution Facility. Secondary communications
room for a building using a star networking topology. The IDF is
dependent on the MDF.

Page 90
 Layer 1 Documentation
 Logical diagram is a snapshot view of over all LAN
implementation
 Useful in troubleshooting problems and implementing
expansion in the future
 Elements of the Logical Diagram Include:
 Exact locations of MDF and IDF's 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.
 Detailed documentation of all cable runs. This is called a 'cut sheet'.

Page 91
 Layer 2
 The purpose of Layer 2 (Data Link Layer) devices in the
network are to provide 'flow control', 'error detection &
correction', and 'reduce congestion' in the network.
 The two most common Layer 2 devices are Bridges and LAN
switches.
 LAN switches can allocate bandwidth on a per port basis
thus allowing more bandwidth to vertical cabling, uplinks
and servers. (Asymmetric switching)

Page 92
 Asymmetric Switching
 Provides the capability to switch between unlike bandwidths.
• E.g. A server could be placed on a 100Mb/s port to reduce potential
bottlenecks, while hosts would be placed on a 10Mb/s port
• Requires the switch to use memory buffering
 Question
• Why does asymmetric switching require a switch to use memory
buffering?

Page 93
 Switching
 Avoid congestion on a LAN by using microsegmentation to
eliminate collision domains
 Cascade switches and hubs
 Tailor availability to the needs of the device

Page 94