Network Topology Architectures

What it is

A network topology architecture is the overall shape of a network: its layers, its links and where traffic flows.

The right shape depends on how big the network is and what it carries.

This topic covers the designs you should be able to recognise and describe:

• two-tier and three-tier campus designs
• spine-leaf data centre design
• WAN
• small office and home office (SOHO)
• on-premises versus cloud

How it works

Each design arranges devices in layers or links to suit its size and traffic pattern.

Three-tier campus design

The classic campus design has three layers.

• Access layer: where end devices connect, such as PCs, phones and access points. Access switches are plentiful and give each user a port.
• Distribution layer: gathers several access switches. It is where policies, VLAN routing and filtering are often applied.
• Core layer: a fast backbone that links the distribution layers and the rest of the network. Its job is to move traffic quickly, so it avoids complex policy.

This design scales well for a large campus with many buildings. Each layer has a clear purpose, which makes the network easier to grow and troubleshoot.

Two-tier (collapsed core) design

In a smaller campus, a separate core is not worth the cost.

The two-tier design merges the core and distribution layers into one layer. This is called a collapsed core.

The access layer connects up to the collapsed core switches. It gives most of the benefits of three tiers with fewer devices.

Spine-leaf design

Data centres move a lot of traffic between servers, which is called east-west traffic.

The three-tier design is built more for traffic going up and out, called north-south. The spine-leaf design suits east-west traffic better.

Every leaf connects to every spine. Leaves do not connect to each other, and spines do not connect to each other.

Servers connect to leaves. Any server reaches any other server in the same number of hops: leaf, spine, leaf.

That gives predictable delay and many equal paths. To add capacity, you add spines. To add ports, you add leaves.

WAN

A wide area network (WAN) connects sites that are far apart, such as a head office and branch offices.

The links are usually provided by a service provider. Common choices include a private line, a metro Ethernet service, an MPLS service, and an encrypted VPN across the internet.

A WAN is typically slower and more expensive per megabit than a LAN, so its design focuses on cost, reliability and security.

SOHO

A small office or home office usually has a small number of users.

A single device often does the work of router, switch, wireless access point and firewall, connected to an internet provider. There is no distribution or core layer because the network is so small.

On-premises and cloud

An on-premises network is built and run by the organisation in its own buildings. It controls the hardware and pays for the equipment and its upkeep.

A cloud network uses computing resources hosted by a provider. The usual service models are:

• IaaS: you rent virtual machines, storage and networks.
• PaaS: you rent a platform to run your own applications.
• SaaS: you use a finished application, such as email.

Cloud can be public (shared provider infrastructure), private (dedicated to one organisation), or hybrid (a mix of on-premises and cloud). Many organisations run a hybrid network and connect the two with a VPN or a private link.