Network Topologies and Segmentation
Zones, boundaries and the perimeter mindset
Topology is how zones are arranged (DMZ, trusted, untrusted), and segmentation is how traffic is confined between them. Both turn 'we have a network' into 'we have a defense in depth'.
What you will learn
- Design a tiered topology with DMZ, trusted and management zones
- Explain how segmentation reduces blast radius
- Trace a user's request through the zones to name where controls sit
What is this?
Network topology describes how devices and zones connect; segmentation describes where traffic is confined and denied. A classic design separates untrusted (internet), semi-trusted (DMZ with public servers), trusted (internal users) and management zones, with firewalls between them. Good segmentation turns a single compromised host into a contained event instead of a network-wide breach.
Why do you need it?
- Flat networks fail fast: one exploited workstation can reach the database fleet wholesale.
- The DMZ pattern lets you expose services without exposing trust.
- Segmentation zoes are where detection value lives: cross-zone traffic should be rare, so any of it is interesting.
- Compliance (PCI, SOC 2) increasingly requires evidenced segmentation.
Mental model
A house with zones and locked doors
A segmented network
The porch (DMZ) faces the street and holds the mail slot; the foyer lets guests in but not upstairs; the office and kids' rooms are zones with rules about who may cross. A burglar in the porch doesn't get the bedroom jewelry.
How it works
- Zones are defined by policy (what runs there) and by ACLs/firewalls (what may cross in/out).
- The DMZ holds servers that must be reachable from the internet; the firewall lets web traffic in but the DMZ cannot initiate back into trusted.
- Trusted zones terminate the user workforce; management zones host admin paths and bastions, reachable only from admin subnets.
- Segmentation is applied with firewalls, VLANs (layer 2 isolation), subnets (layer 3) and, increasingly, identity-aware policy (zero trust).
- Every cross-zone flow should be permitted by exception and observable.
A small segmented topology.
In production
- A web stack lives in the DMZ; the database cluster lives in a dedicated DB zone the DMZ can reach only on 5432.
- Workstations cannot initiate to the management zone at all — admins come in via a bastion with MFA.
- A SIEM alert fires on 'DMZ → trusted zone' flows because they should only occur as documented exceptions.
Best practices
- Default-deny between zones; document every exception rule with an owner and expiry.
- Keep the DMZ minimal and patched — it's the most exposed real estate.
- Route management traffic on a dedicated plane and isolate it from user/data traffic.
- Test segmentation continuously: probe each boundary from authorized lab hosts so theoretical isolation is proven isolation.
Common mistakes
- Designing zones but deploying a flat network 'for now' — 'for now' becomes the breach.
- Putting the database in the DMZ to 'simplify' the firewall.
- Forgetting management-plane isolation, so RDP to the domain controller rides the user VLAN.
- Segmentation without monitoring: if no one watches the boundaries, a wider open door may never be noticed.
Debugging tips
- When a zone 'shouldn't reach' another but does, check for extra routes, trunk misconfig and duplicate subnets.
- Trace with traceroute across each boundary; a control that isn't the named firewall is a red flag.
- Query the firewall's session table for flows between zones and review exceptions.
Security considerations
- Segmentation is only as strong as the boundary device's own hardening — a public web server with admin interfaces exposed is a DMZ in name only.
- Zero-trust architecture builds on segmentation by adding per-workload, identity-based micro-segmentation where flattening is unavoidable.
Interview checkpoint
Quick quiz
What is the primary purpose of a DMZ?
Flat network design maximizes the blast radius of any single compromise.