Network topology is the arrangement of a network’s parts, the sites, devices and links, and the paths traffic takes between them. In a local network it describes how switches, access points and devices connect. In a wide area network (WAN) it describes how offices, data centers and cloud connections link together, for example all through a central hub or directly to one another. Topology shapes performance, cost and what happens when a link or site fails.
At a glance
- Topology is the map of how a network’s sites, devices and links connect, and how traffic flows over them.
- Common WAN designs are hub-and-spoke, full mesh and partial mesh; many real networks mix them.
- Physical topology (cables and circuits) and logical topology (traffic paths and tunnels) can differ.
- Topology determines where single points of failure sit and how far traffic travels between sites.
- Changes such as moving applications to the cloud often make an existing topology a poor fit.
What problem it solves
Every network has a topology, whether it was designed or grew by accident. Knowing it, and choosing it deliberately, answers practical questions: if this site or circuit fails, who loses access? Why is traffic between two branches slow? Where should a new data center or cloud connection plug in? Which links need to be larger?
A poorly matched topology shows up as slow applications, outages that spread further than they should, and circuits that cost more than they need to. For example, sending all internet and cloud traffic through headquarters made sense when applications lived there, but now often adds delay and loads the headquarters circuit. Designing the topology around where users and applications actually are addresses the cause rather than the symptoms.
How it works
Hub-and-spoke (star). Each site connects to a central hub, and traffic between sites passes through it. This is simple and economical to manage, and it suits traffic that flows mostly to one place. The hub is a potential single point of failure unless it is built with redundancy. See hub-and-spoke network.
Full mesh. Every site has a direct path to every other. Built with dedicated physical links, the link count grows with every pair of sites, so a physical full mesh is usually reserved for small groups of critical sites, such as data centers. Built with overlay tunnels, a logical full mesh can give any-to-any connectivity over a much smaller set of access circuits, often one or two per site. Either way, traffic between sites does not have to detour through a hub.
Partial mesh. Some sites connect directly, others through hubs. This is common in practice: regional hubs, direct links between busy sites, and spokes for small offices.
Ring and bus. Rings, where each node connects to two neighbors, appear in carrier fiber and some campus and industrial networks because traffic can go either way around if a link fails. Bus designs are now largely historical in office networking.
Physical versus logical. Overlays such as SD-WAN can build a logical mesh of tunnels on top of a physical network where each site has only one or two circuits. Two designs that look different on a diagram may share the same underlying fiber, so diversity needs checking at the physical level.
A network assessment is often the first step in documenting a topology. Our Managed Network Services solution page covers providers that design and run it.
When it matters for buyers
- Before buying SD-WAN or new circuits. Decide which sites should talk directly, which should use hubs and where cloud traffic should leave the network.
- After an outage. If one failure took down more than expected, the topology probably has a hidden single point of failure.
- Cloud migration. Moving applications out of a data center changes traffic patterns, often away from hub-and-spoke.
- Mergers and new sites. Joining networks means choosing a combined design instead of stacking two.
- Resilience planning. Redundancy depends on topology: a second circuit adds the most protection when it follows a different physical path, and ideally lands at a different place in the network.
Questions to ask vendors
- What topology do you recommend for our sites, and why for our traffic patterns?
- Where are the single points of failure in that design, and what does it cost to remove each one?
- How will cloud and internet traffic leave the network: centrally or at each site?
- Can you show physical paths, not just logical tunnels, for our redundant links?
- How does the design scale as we add sites?
- Will you provide and maintain current topology diagrams?
How it differs from network segmentation
Topology is about how the network is physically and logically connected: which sites and devices link to which, and the paths between them. Network segmentation is about dividing the network into zones and controlling what traffic may pass between them, mainly for security. A flat network with a simple star topology can be heavily segmented, and a complex mesh can be almost unsegmented. Both appear on network diagrams, so they are often discussed together, but they answer different questions: topology asks how traffic can get there, and segmentation asks whether it is allowed to.