Backbone Network Design Network Design And Performance Analysis.pdf

sunalisharma 46 views 12 slides Aug 03, 2024
Slide 1
Slide 1 of 12
Slide 1
1
Slide 2
2
Slide 3
3
Slide 4
4
Slide 5
5
Slide 6
6
Slide 7
7
Slide 8
8
Slide 9
9
Slide 10
10
Slide 11
11
Slide 12
12

About This Presentation

Network Design And Performance Analysis


Slide Content

EE5302 Network Design and Management 1
Backbone Network Design
EE5302 Network Design and Management2
Dr. W Yao, Brunel University
Backbone Network Design

The decision must be made whether to interconnect the
access design, build a backbone network, or interface to
an existing backbone such as a public data network
service like IP and frame relay.

The capacity required to support the access network
was determined primarily from the outputs of the design
stage.

A topology and style consistent with the services and
technology must be chosen.

The resultant topology will in turn determine the number
and placement of backbone nodes, the circuit design,
and the overall integrated access and backbone network
design.
EE5302 Network Design and Management3
Dr. W Yao, Brunel University
Backbone Network Design

Backbone Requirements

Backbones provide many efficiencies not achievable
from a meshed-access network, including

Traffic consolidation – elimination of multiple paths with
different type of traffic

High-bandwidth switched-services platform

Rerouting and redundancy

Economies of scale

Self-healing architecture

Sharing of equipment and facilities by multiple locations

Intelligent routing

Dynamic bandwidth resource allocation

Flexible topologies and styles of design

Distributed or centralized network management

Flexibility
EE5302 Network Design and Management4
Dr. W Yao, Brunel University
Backbone Network Design
Private Line WAN Backbone Compared to Switched Service WAN Backbone

EE5302 Network Design and Management5
Dr. W Yao, Brunel University
Backbone Network Design

As the number of point-to-point trunks grows between
access nodes, it becomes necessary to design a
backbone network layer.

Public switched-data networks can also act as
backbone networks in a hybrid fashion to an access
network.

When viewed from the backbone network, the
perspective must be global.

Interfaces

The primary interfaces for the backbone design will be either
access circuits from the access design portion or direct-user
access into the public service backbone.

The primary speeds for backbone interfaces are 56 kbps, FT1,
DS1, DS3, 100 Mbps, 155 Mbps, and OC-N.
EE5302 Network Design and Management6
Dr. W Yao, Brunel University
Backbone Network Design

Protocols

Many of the user protocols will be transparent to the
backbone, but the design of the backbone may still have an
effect on them.

The backbone design must be flexible enough to
accommodate multiple protocols and operating systems,
whether switching them transparently or actually becoming
involved in the intelligent operations of the protocol.

Backbones must have the capability to interconnect diverse
sites, switch or route all the protocols of the access network,
and accommodate both in-band and out-of-band protocol
transmissions.

TCP/IP remains the most common backbone network and
internetwork protocol, and can be deployed in more ways.

Frame relay is one of the most popular publicly-used
backbone protocols.
EE5302 Network Design and Management7
Dr. W Yao, Brunel University
Backbone Network Design

Take into account the protocols passed over the backbone,
even though they may be passed transparently.

Routable protocols such as TCP/IP are the preferred WAN
protocols, while protocols such as LAT, IBM SRT, and
NetBIOS (NetBEUI) can only be bridged.

Determine whether the protocols are operating half or full
duplex.

Half-duplex protocols generate large amounts of overhead
because the sender has to wait for acknowledgements
from the receiving station before more information can be
sent. Thus, over half of the available bandwidth is wasted
with turn-arounds.

Always minimize and localize the routing tables in each router.

The backbone designs should always attempt to use standard
protocols for wide area networking.
EE5302 Network Design and Management8
Dr. W Yao, Brunel University
Backbone Network Design

Architecture and Technology

The network backbone is either the same or one technology
further advanced than the access network technology.

IP, by far the most common backbone technology, is
usually ubiquitous across the LAN and WAN.

Technologies such as ATM are offering a common
platform and architecture for the transport and
internetworking of almost all existing protocols and traffic
types (i.e. private line, FR, IP, and SMDS services over an
ATM backbone).

Backbone designs should also be faster than the access
devices, at least on a maximum burst, throughput, or
utilization basis.

Otherwise, performance would degrade, file transfer
would take longer over the WAN, bottlenecks would occur.

EE5302 Network Design and Management9
Dr. W Yao, Brunel University
Backbone Network Design
Access and Backbone Network Layer Migration Concept
EE5302 Network Design and Management10
Dr. W Yao, Brunel University
Backbone Network Design

Long-term planning is easier with the backbone than with
access, because capacity additions and technology changes
can be more easily migrated into the network in layers.

Access, configuration, protocol, service and technology
changes require replacement of core switching equipment.

Determine whether to build a connectionless or connection-
oriented service and then determine their impact on each other
if layered.

Connectionless and connection-oriented services each offer
their own advantages in efficiency and throughput,
depending on data unit sizes, tr affic type (voice, data, and
video), node-to-node overhead involved, and the delays
inherent in setting up a connection.

The primary concern is optimizing the packet, frame, or
cell size to ensure that the amount of overhead generated
is at a minimum in order to guarantee the required quality
of service.
EE5302 Network Design and Management11
Dr. W Yao, Brunel University
Backbone Network Design

Features, Functions, and Services

The real benefits derived from using a network backbone are
the public-network value-added, switched-services available.

These services include circuit switching, packet-switching,
TCP/IP, FR, SMDS, and ATM.

It is less expensive for the user to simply access them
rather than build their own corporate backbone.

These service backbones must also support many of the
features and functions that operate on the access devices.

Make sure devices follow industry standard protocols to
assure Customer Premises Equipment (CPE) compatibility.

Issues like congestion control in multiprotocol environments
must be planned for and contingency plans developed.

Understand the deficiencies of each service offered on the
backbone and make sure the customer and user also
understand them.
EE5302 Network Design and Management12
Dr. W Yao, Brunel University
Backbone Network Design

Backbone Network Capacity Required „
Once the access loading is determined, the total backbone
capacity can be found.

Backbone capacity is typically measured by ingress and
egress bandwidth, backbone bandwidth between switches,
processing power, and port density of each switch.

Node type is selected based on the amount of local,
remote, and pass-through traffic processing, along with
the service type required.

It is then up to the designer to apply good judgment (and
a good design tool) for each backbone node design – this
ensures that the loading can handle both normal operating
conditions and a single (or multiple) node or link failure
analysis.

EE5302 Network Design and Management13
Dr. W Yao, Brunel University
Backbone Network Design „
Backbone Node Selection

Determine what percentage of the traffic will remain
in the access design and what needs to be passed
over to the backbone.

Perform a “backbone traffic matrix” in the same manner as
the access matrix, consider the access-node traffic (instead
of user-application traffic) passed to the backbone.

Select a starting location for each backbone node you think
that may be required.

Determine what percentage of traffic passed to each
backbone node goes in and back out of the same
backbone node.

Determine how much traffic enters the backbone node and
leaves to go to another backbone node.
EE5302 Network Design and Management14
Dr. W Yao, Brunel University
Backbone Network Design
Typical Traffic Patterns – Single-Trunk Access Nodes
EE5302 Network Design and Management15
Dr. W Yao, Brunel University
Backbone Network Design
Typical Traffic Patterns – Dual-Trunked Access Nodes
EE5302 Network Design and Management16
Dr. W Yao, Brunel University
Backbone Network Design

Utilization, Loading Factors, and Anticipating
Failures

The same calculations for the access network can now be used
for the backbone network to calculate node and link utilization
and loading.

These calculations will be more precise because the traffic
patterns between the access nodes are already (accurately)
known or approximated, and thus the backbone nodal design
will be more accurate.

It is easier to change the utilization and loading of the
backbone if required. And it is easier to throw more capacity
(bandwidth or circuits) at the backbone.

Performance problems can be isolated and fixed more easily on
a backbone network than on the access side, where user traffic
can be temporarily rerouted around the “construction area”.

EE5302 Network Design and Management17
Dr. W Yao, Brunel University
Backbone Network Design

Major Traffic Patterns

Most or all traffic that enters a node leaves the same node
and does not transit any other backbone nodes. The backbone
nodal trunks are primarily used for backup and redundancy.

Traffic originating on a back bone node is transmitted
symmetrically to every other backbone node. This is the case
for broadcast or public network. The backbone nodal trunks
are primarily used for switchin g and the links between them
are heavily used.

All traffic patterns are asymmetric al and are divided into user
classes such as terminal-to-host and LAN-to-LAN (IP)
communications. The backbone nodal trunks are primarily
used for switching or routing and the link usage varies.

Users never talk to nodes on the same backbone (this is a
multiple backbone scenario with backbone nodes connected
via WAN links). The applications for this case are varied, but
this case relates to a public-network service.
EE5302 Network Design and Management18
Dr. W Yao, Brunel University
Backbone Network Design
Backbone with 100 percent Traffic Remaining with Access BB Node
EE5302 Network Design and Management19
Dr. W Yao, Brunel University
Backbone Network Design
Backbone with 0 percent Traffic Remaining with Access BB Node
EE5302 Network Design and Management20
Dr. W Yao, Brunel University
Backbone Network Design

Total Backbone Capacity

Capacity is measured by the required amount of ports for all
access devices, trunks to other backbone nodes, and the
bandwidth on the trunks between those ports.

Given the backbone-access traffic patterns, backbone capacity
can be calculated in two ways.

Given the number of nodes required, calculate the total
capacity of the backbone network.

Given the capacity required, cal culate the total number of
backbone nodes needed.

As the traffic patterns become more distributed, the network
capacity decreases. This is because there are more options for
traffic to use the limited bandwidth resources.

For this type of network, the only way to add extra
capacity is to overengineer from the beginning.

EE5302 Network Design and Management21
Dr. W Yao, Brunel University
Backbone Network Design

Route Determination

WAN elements such as circuit/packet/frame/cell switches;
bridges; routers; and other associated devices can all route or
switch data based on preexisting physical routes over access
and backbone links.

In IP the exact route is dynamic and not predefined.

In FR these routes are either static or dynamic, and either
preprogrammed (PVC) or assigned dynamically (SVC).

Some are routed “best way” as long as they get to the
required destination. This choi ce of routing is accomplished
node-to-node, hop-by-hop, and is based on a variety of
variables (hop count, cost, bandwidth, priority, quality of line).

The priority of traffic flow may be (from low to high)
copper, coax, short-range microwave radio, VSAT,
wireless communications, or fiber optics.

Do not let the users control the routing on the backbone
network. They could misuse the parameters.
EE5302 Network Design and Management22
Dr. W Yao, Brunel University
Backbone Network Design

Future Capacity

User requirements for bandwidth can increase from 25 to 250
percent on the average annually, and therefore extra capacity
should be built into the backbone.

Loading factors should be low early in the life of the network,
and the network should have the capability to quickly add
capacity when and where required.

This means both increased processing power and extra
trunk bandwidth at each backbone site.

When turning up a new service, make sure that this extra
capacity is available.

Then plan on providing more than what is required based on
the growth rate.

If you follow the design practice of building a backbone with
a higher level of technology and larger bandwidth pipes than
the access actually requires, the design will prove effective.
EE5302 Network Design and Management23
Dr. W Yao, Brunel University
Backbone Network Design

Styles of Topologies

The backbone network topology should be a function of the
user applications, the access-network topology, traffic
volume, and range and profile of connectivity.

Star (hub-and-spoke)

In starstyle, there is a central node serving as the hub node
and all other nodes are connected via point-to-point circuits to
the central node. At least N-1 links are needed for N nodes.

All communications pass through the central node.

This star style is often used in an environment such as a LAN
hub or ATM switch/hub.

While this configuration provides a maximum of two hops, it is
unreliable and susceptible to an entire network failure when
the hub node fails.

Distributed staris typically used in LAN environments that
use hubs as concentrators and tie the hubs together.
EE5302 Network Design and Management24
Dr. W Yao, Brunel University
Backbone Network Design
Star Backbone Topology

EE5302 Network Design and Management25
Dr. W Yao, Brunel University
Backbone Network Design
Distributed Star Backbone Topology
EE5302 Network Design and Management26
Dr. W Yao, Brunel University
Backbone Network Design

Loop

Each network node is connected to two other network nodes.
At least N links are needed for N nodes.

This loop style is often used for distributed networks where
nodes primarily talk to local nodes or point-to-point
communications are required over short distances.

There is no maximum to the number of hops across this
network, but it is reliable up to the point of two link failures,
which would then separate the network into two pieces.

Capacity planning is difficult with loop topologies, and
upgrades are difficult if the traffic patterns are not symmetric
and consistent across different access nodes.
EE5302 Network Design and Management27
Dr. W Yao, Brunel University
Backbone Network Design
Loop Backbone Topology
EE5302 Network Design and Management28
Dr. W Yao, Brunel University
Backbone Network Design

Meshed and Fully Meshed

The degree to which a mesh is built depends upon the hardware
and software expense of the ports, incremental cost of the links,
network resource availability, an d the amount of reliability and
redundancy required.

The number of links required for a fully meshed design (N nodes)
is N(N-1)/2.

Daisy-Chained Access Nodes

All network-access devices are dual-homed to two high-capacity
backbone switches.

This provides for high availab ility, but wastes bandwidth if
the applications are regional or th eir processing is distributed.

Each access device also acts as a switch through a daisy chain.

The connectivity requirements are retained. The number of
required links, equipment cost, distance between nodes, and
reliability are reduced.

EE5302 Network Design and Management29
Dr. W Yao, Brunel University
Backbone Network Design
Semimeshed and Fully Meshed Backbone Topology
EE5302 Network Design and Management30
Dr. W Yao, Brunel University
Backbone Network Design
Daisy Chain Access to Backbone Network
EE5302 Network Design and Management31
Dr. W Yao, Brunel University
Backbone Network Design

Backbones within Backbones

In many cases, the backbone nodes and their topology are
separated from the access-node topology. For example, many
network providers often build multiple backbones within
backbones in a hierarchical nature.

Networks with hierarchical structures offer not only
redundancy and availability, but also a reduced hop count
(and the processing delay that goes along with it) and an
increased speed of traffic that transits long distances
across the network.
EE5302 Network Design and Management32
Dr. W Yao, Brunel University
Backbone Network Design
Single-Access and Single-backbone Network Configuration

EE5302 Network Design and Management33
Dr. W Yao, Brunel University
Backbone Network Design
Single-Access and Multiple-Backbone Network Configuration
EE5302 Network Design and Management34
Dr. W Yao, Brunel University
Backbone Network Design

Backbone Topology Strategies

The enterprise backbone provides not only
interconnection of the departmental subnets, but also
LAN/MAN/WAN interconnectivity through routers or
switched networks (e.g. IP, SMDS, and ATM).

Desirable Topologies per Technology

With constant bit-rate traffic, the WAN connection of choice is
often a dedicated or switched circuit.

For traffic of a variable bit-rate nature, WAN technologies
include packet switching, point-to-point solutions, IP, FR,
SMDS, ATM, and switched-data services.

Both types of connections may use switches, bridges, routers,
multiplexers, and a variety of other access hardware.
EE5302 Network Design and Management35
Dr. W Yao, Brunel University
Backbone Network Design

Combining the right topology with the right technology, and
vice versa, will avoid design flaw s. Make sure the device so
designed grows with the network , rather than limiting it
through proprietary implementations of older technologies.

Some devices can exceed the physical limitations of DS1 and
DS3 circuits, but only in a local (limited distance) area.

Routers can use HSSI, FDDI, and SONET interfaces, which
allow allocation of bandwidth in Mbps and Gbps chunks,
rather than be constrained to DS1 speed.

Some devices may be more desirable than others because of
their value-added services, while others may save expenses.

Using bridges as distance extenders rather than running
coax cable.

For packet-switched technologies, the topologies available are
quite flexible (e.g. IP and FR).

The integrated circuit- and packet -switched equipment provides
good integration of packet- and circuit-switching capabilities.
EE5302 Network Design and Management36
Dr. W Yao, Brunel University
Backbone Network Design

Requirements Drive the Topology

Shortest Path Design Methodology

Add the links that are absolutely required first, and
proceed to add more links until all possible links are
added where capacity is required for data flow.

Eliminates links and combines traffic over other existing
links based on many factors such as shortest path, link
cost, and quality of facilities.

Always review all technologies and routing algorithms
available, and do not confine yourself to a single transport
technology or protocol during the design process.

Never lose sight of the original user requirements during the
backbone design.

Understanding the effects that your backbone design will
have on user applications is critical in optimizing the
backbone design. This is a two way process.

EE5302 Network Design and Management37
Dr. W Yao, Brunel University
Backbone Network Design
Shortest Path Design Methodology
EE5302 Network Design and Management38
Dr. W Yao, Brunel University
Backbone Network Design

Hybrid Topologies

Hybrid networks that employ both fixed and switched
topologies are common. Switched services are often used as
backup to leased facilities.

If designing a private network, make sure it can be integrated
(hybridized) with a public network-switched services.

Public and private network hybrids allow control of critical
resources to be retained while taking advantage of public
services functionality and pricing.

When designing a hybrid backbone network,

Attempt to keep all speeds fairly constant, and beware of
bottleneck bandwidth points.

Beware of multiple encapsulation schemes in hybrid
networks, and understand that similar network devices
must often be used at network ingress and egress points.
EE5302 Network Design and Management39
Dr. W Yao, Brunel University
Backbone Network Design

Topology of the Future – Distributed Design

The primary feature of distribute d processing networks is the
placement of the intelligence to process and switch data into
the access devices, rather than the backbone devices.
EE5302 Network Design and Management40
Dr. W Yao, Brunel University
Backbone Network Design

Network Management

Network management is much easier to administer at
the network level than at the access concentrator or
even user-device level.

A single WAN management platform could manage
the addressing, bandwidth-management (e.g.
congestion control, throughput classes,
synchronization), administration, and performance-
management activities.

EE5302 Network Design and Management41
Dr. W Yao, Brunel University
Backbone Network Design

Total Network Timing

Timing is always an important aspect of each design with any
technology whose success relies on error-free synchronization.

Timing is also important between the Customer Premises
Equipment (CPE) and network-access devices.

The higher the transmission speed, the more critical timing
becomes.

External timing source should be used when possible.

Always use a clock source as accurate as possible.

Analyze time sources on a regular basis and always look for
potential timing loops (a device that provides a clock source and
receives back its same clock as its own source).

Timing problems can cause line interference, data unit slips
which lead to loss of data, interruption of service, and a general
decrease in the reliability of transmission.
EE5302 Network Design and Management42
Dr. W Yao, Brunel University
Backbone Network Design

Tuning the Network

Network tuning should be performed at the same time
you are designing the backbone network, as well as at
specific intervals in the lifetime of the network.

Four specific areas, when tuned, will increase the
efficiency and throughput of both the access and
backbone design.

Optimizing packet, frame, and cell sizing

Limiting the segmentation of user data by lower-level
protocols

Decreasing the overall port-to-port transfer delay

Using window size changes to flag potential network
degradation
EE5302 Network Design and Management43
Dr. W Yao, Brunel University
Backbone Network Design

Optimizing Packet/Frame/Cell Size

There is a tradeoff between large and small packet sizes.

Small packets
: better response time, less data corruption
due to errors (less retransmissions), but more overhead.

Large packets
: ……

Packet size tuning is a balancing act, where additional factors
influencing the size include:

How long it takes to read in a packet

Buffer space taken by held packets

Packets per second (pps) dropped

Time to forward packet to next device or destination

Mix of protocols to bridge/route

The best method for achieving optimal packet size is to tune
the network based on the performance observed and refer to
a trending analysis or design tool for accuracy.
EE5302 Network Design and Management44
Dr. W Yao, Brunel University
Backbone Network Design
Idea Packet Size Curve

EE5302 Network Design and Management45
Dr. W Yao, Brunel University
Backbone Network Design

Limiting Protocol Segmentation

Try to reduce the amount of segments at both the user file-
transfer level and at the acce ss and backbone-technology
transport level.

When many layers of encapsulation are encountered across
the network, packets may have to be broken down at each
new protocol level and decrease throughput.

Port-to-Port Data Transfer Delay

The port-to-port data transfer delay becomes more
appreciable the larger the file size.

Much of this delay and overhead depends on internal
architecture and software protocol handling of the device.

Many devices implement multiple ports per interface card,
thus reducing this effect.

The higher the bus speed, the faster the data rate between
interface boards and the central (or distributed) processor(s).
EE5302 Network Design and Management46
Dr. W Yao, Brunel University
Backbone Network Design

Window Sizes

The window size determines how many packets can be
outstanding in the network before an acknowledgement is
received from the last unacknowledged packet sent.

Window size can be tuned at each level of the X.25 packet-
switched network, or at the network layer (TCP) portion of
the protocol stack (usually in the user or access device).

These windows provide transmission-flow control.

TCP will adjust its window size automatically based on the
current network throughput.

Increase window size provides increased throughput, but
requires more memory and buffers in the network.

Network performance problems can be determined by
trending the window size changes and determining where
additional throughput may be required in the network.
EE5302 Network Design and Management47
Dr. W Yao, Brunel University
Backbone Network Design

Bursting

Most protocols are designed to allow the application the
capability to burst above a predefin ed traffic parameter setting.

Many applications naturally sent bursts of data over dedicated
facilities when they have the entire bandwidth.

Statistical multiplexers allow a form of bursting.

Two measures of bursts are the committed burst size for
forward and backward traffic and the excess burst size for
forward and backward.

Frame relay was designed to allow network access for multiple
variable-bit rate users.

These are the same users who once paid for low-speed
private lines to connect local computing and the
communication environment.

While users can intelligently use the CIR and maximize
throughput, their savings are passed off to other users in
the form of delay.
EE5302 Network Design and Management48
Dr. W Yao, Brunel University
Backbone Network Design
Effect of Frame Relay Service CIR on User Traffic