Describe Characteristics of Network Topology Architectures

Every network — from a small home office to a global cloud provider’s data center — is built around some underlying topology: a pattern describing how devices are physically or logically connected to one another. Choosing the right topology affects everything: cost, redundancy, scalability, performance, and how easy the network is to troubleshoot.

This article walks through the major network topology architectures you’ll encounter in real networks and on the CCNA exam: two-tier (collapsed core), three-tier, spine-leaf, along with cloud and WAN architecture concepts like on-premises, cloud, and hybrid designs, and SDA (Software-Defined Access).

What Is Network Topology?

A topology describes the arrangement of network devices and the links between them. There are two perspectives:

• Physical topology — how cables and devices are physically arranged.
• Logical topology — how traffic actually flows, regardless of physical layout (which may differ due to VLANs, tunneling, or virtualization).

Traditional Three-Tier Hierarchical Architecture

This is the classic Cisco-recommended enterprise campus design, consisting of three layers:

graph TB
    subgraph Core["Core Layer"]
        C1["Core Switch 1"]
        C2["Core Switch 2"]
    end
    subgraph Distribution["Distribution Layer"]
        D1["Dist Switch 1"]
        D2["Dist Switch 2"]
    end
    subgraph Access["Access Layer"]
        A1["Access SW 1"]
        A2["Access SW 2"]
        A3["Access SW 3"]
    end
    C1 --- D1
    C1 --- D2
    C2 --- D1
    C2 --- D2
    D1 --- A1
    D1 --- A2
    D2 --- A2
    D2 --- A3

The Three Layers Explained

Characteristics

• Highly redundant — every access switch connects to at least two distribution switches; every distribution switch connects to at least two core switches.
• Scalable — new access-layer “blocks” can be added without redesigning the whole network.
• Clear separation of duties — policy enforcement happens at distribution, not at the core, keeping the core layer extremely fast and simple.
• More hardware and cabling cost compared to simpler designs — appropriate for larger campuses.

Two-Tier (Collapsed Core) Architecture

In smaller networks, the distribution and core layers are collapsed into a single layer, reducing cost and complexity while sacrificing some scalability.

graph TB
    subgraph CollapsedCore["Collapsed Core/Distribution Layer"]
        CD1["Switch 1"]
        CD2["Switch 2"]
    end
    subgraph Access["Access Layer"]
        A1["Access SW 1"]
        A2["Access SW 2"]
    end
    CD1 --- CD2
    CD1 --- A1
    CD1 --- A2
    CD2 --- A1
    CD2 --- A2

Characteristics

• Lower cost — fewer devices and interconnections needed.
• Simpler to manage — fewer layers means fewer places to troubleshoot.
• Best suited for small-to-medium networks — a single building or small campus.
• Less scalable — as the network grows, this design can become a bottleneck, eventually requiring migration to a full three-tier design.

Spine-Leaf Architecture (Data Center)

Spine-leaf is the dominant modern data center topology, designed for the very different traffic patterns found in data centers, where most traffic flows server-to-server (east-west) rather than client-to-server (north-south) as in traditional campus networks.

graph TB
    subgraph Spine["Spine Layer"]
        S1["Spine 1"]
        S2["Spine 2"]
    end
    subgraph Leaf["Leaf Layer"]
        L1["Leaf 1"]
        L2["Leaf 2"]
        L3["Leaf 3"]
    end
    S1 --- L1
    S1 --- L2
    S1 --- L3
    S2 --- L1
    S2 --- L2
    S2 --- L3

Key Characteristics

• Every leaf switch connects to every spine switch — this is the defining feature. There are no leaf-to-leaf or spine-to-spine direct connections.
• Consistent, predictable latency — any server-to-server path is always exactly two hops (leaf → spine → leaf), regardless of which servers are communicating.
• No collapsed core/distribution concept — spine switches perform pure high-speed switching; leaf switches connect to servers/hosts.
• Highly scalable horizontally — need more capacity? Add more spine switches (increasing bandwidth) or more leaf switches (increasing port count), without redesigning the whole fabric.
• Uses Layer 3 routing between leaf and spine in modern designs (often with ECMP — Equal-Cost Multi-Path — to use all links simultaneously, unlike STP-based designs which block redundant links).

Comparison Table: Three-Tier vs. Two-Tier vs. Spine-Leaf

On-Premises, Cloud, and Hybrid Architectures

Beyond physical topology shapes, modern network design also considers where infrastructure lives.

On-Premises

All infrastructure (servers, switches, routers, firewalls) is owned, hosted, and managed within the organization’s own physical facilities.

• Pros: full control, no recurring cloud costs, data stays within owned facilities (useful for certain compliance requirements).
• Cons: high upfront capital expense, the organization is responsible for all maintenance, cooling, power, and hardware refresh cycles.

Cloud

Infrastructure and services run on a third-party provider’s platform (AWS, Azure, Google Cloud), accessed over the Internet or private connections.

• Pros: elastic scalability, pay-as-you-go pricing, provider handles physical maintenance.
• Cons: ongoing operational cost, dependency on provider reliability and pricing model, potential data residency/compliance concerns.

Hybrid

A combination of on-premises and cloud resources, often connected via a dedicated private link (e.g., AWS Direct Connect, Azure ExpressRoute) or a secure VPN tunnel.

• Pros: flexibility to keep sensitive workloads on-premises while bursting to cloud for scalability; gradual cloud migration path.
• Cons: increased architectural complexity; requires careful design of routing, security policy, and identity management across both environments.

graph LR
    OnPrem["On-Premises Data Center"] ---|"VPN / Direct Connect / ExpressRoute"| Cloud["Public Cloud (AWS/Azure/GCP)"]
    Users["Remote Users"] --> Cloud
    Users --> OnPrem

Software-Defined Access (SDA)

SDA is Cisco’s software-defined networking architecture for campus networks, built on top of a physical spine-leaf-like fabric. It separates the network into:

• Underlay — the physical, traditional IP routing infrastructure providing basic reachability (often built as a simplified two-tier or spine-leaf design).
• Overlay (Fabric) — a logical layer built using VXLAN encapsulation, providing segmentation (via VRFs and Scalable Group Tags) independent of the physical topology and IP addressing underneath.
• Controller (DNA Center) — centralized management, policy definition, and automation.

Characteristics

• Decouples logical network segmentation from physical IP addressing — devices can be grouped and policed by identity/group rather than IP subnet.
• Enables consistent policy enforcement regardless of where a device physically connects.
• Significantly reduces manual, box-by-box CLI configuration through centralized automation.

WAN Topology Types

At the wide-area network level, several classic topology shapes describe how remote sites connect:

graph TB
    subgraph HubSpoke["Hub-and-Spoke"]
        H["Hub Site"]
        S1["Spoke 1"]
        S2["Spoke 2"]
        S3["Spoke 3"]
        H --- S1
        H --- S2
        H --- S3
    end

Best Practices for Choosing a Topology

1. Match the topology to the traffic pattern. Client-server-heavy campus networks fit three-tier/two-tier designs; server-to-server-heavy data centers fit spine-leaf.
2. Design for the future, not just today. A two-tier network that works today may need to evolve into three-tier as the organization grows — plan expansion points in advance.
3. Always build in redundancy at every layer where budget allows — single points of failure (a single core switch, a single WAN hub) undermine the whole design.
4. Consider operational complexity, not just technical elegance. A full-mesh WAN might be technically ideal but operationally and financially impractical for many organizations — partial mesh or hub-and-spoke may be the pragmatic choice.
5. Evaluate hybrid cloud connectivity requirements early — retrofitting secure, reliable cloud connectivity into an existing on-premises design is harder than planning for it from the start.

Troubleshooting Topology-Related Issues

• Asymmetric or unpredictable latency in a data center → check whether the fabric is behaving like true spine-leaf (equal-cost paths) or if STP/legacy blocking is limiting available paths.
• Single point of failure discovered during an outage → review the topology diagram for any layer/site with only one active link or one active device; add redundancy.
• Spoke-to-spoke performance issues in hub-and-spoke WAN → confirm whether traffic is being forced through the hub even when a more direct path might be justified by traffic volume — may indicate a need to move to partial mesh for that specific site pair.
• Policy inconsistency across campus locations → may indicate a need to migrate toward an SDA fabric where policy is centrally defined and consistently enforced regardless of physical attachment point.

Summary

Network topology architecture is about matching the shape of your network to the shape of your traffic and your organization’s growth plans. Three-tier designs suit large, hierarchical campus networks; two-tier (collapsed core) suits smaller ones; spine-leaf suits data centers with heavy east-west traffic; and modern approaches like SDA and hybrid cloud connectivity add software-defined flexibility and multi-environment integration on top of these physical foundations. Understanding the strengths and trade-offs of each architecture is essential for designing networks that are both performant today and adaptable tomorrow.

Further Reading

• Cisco: Campus Network Design Guide
• Cisco: Spine-Leaf Architecture Overview
• Cisco: Software-Defined Access (SDA) Overview
• Cisco: Hierarchical Network Design Overview
• AWS: Hybrid Cloud Architecture Patterns