A VLAN Is Not a Firewall: Getting Network Segmentation Right
Network segmentation is frequently presented as a relatively simple security measure. Create separate VLANs for management, production, storage and customer systems, and the infrastructure is now segmented. That's a useful starting point, but it doesn't necessarily mean those networks are securely isolated. A VLAN provides Layer 2 separation. Whether traffic can move between those networks depends on the routing and security controls around them. And that distinction matters considerably.
Separation doesn't automatically mean restriction
Imagine an infrastructure environment with three networks. One contains production virtual machines, another contains management interfaces, and the third provides access to backup infrastructure. All three are assigned separate VLANs. So far, so good. Now suppose a Layer 3 switch routes traffic freely between them. A compromised production machine may be able to communicate directly with management interfaces or backup services, provided no other controls prevent it. The networks are separated at Layer 2, but they aren't necessarily protected from one another.
Network segmentation becomes an effective security control when traffic between security zones is deliberately restricted. That might involve stateful firewalls, router access control lists, distributed firewall policies or other enforcement mechanisms. The particular technology matters less than having clear rules about which systems should communicate and why.
Management access deserves special treatment
One of the more significant risks in virtualised infrastructure is unrestricted access to management systems. Hypervisor administration, storage controllers, switch management, backup consoles and out-of-band interfaces provide considerable control over an environment. These interfaces shouldn't be exposed to ordinary production workloads simply because they happen to exist within the same datacentre. Dedicated management networks, restricted administrative access, strong authentication and controlled routing all reduce unnecessary exposure.
Out-of-band management is particularly valuable when the production network becomes unavailable. However, an isolated management interface is only useful if administrators can still reach it through an appropriate, protected access path during an incident.
Segmentation can also break things
There's another side to this discussion. Applying restrictive firewall policies without understanding application dependencies can cause serious operational problems. Backup software might need access to hypervisor APIs. Monitoring platforms may require connectivity to management services. Cluster nodes might exchange traffic using protocols that aren't immediately obvious from an application diagram. A firewall rule that looks sensible in isolation might interrupt an important infrastructure function.
This is why network segmentation should begin with understanding actual communication requirements. Document the traffic flows, establish the trust boundaries and implement restrictions in a controlled manner. Then test them. Both allowed and prohibited communication should be verified. A policy that permits required services but unintentionally allows everything else isn't particularly useful. The same principle applies to IPv6; introducing dual-stack networking without equivalent security controls can create paths that were never considered in an IPv4-only design.
At KORE, network architecture forms part of the wider security and infrastructure design process. Virtualisation, private cloud, customer connectivity and management systems all depend on correctly designed network boundaries. Good segmentation reduces the potential impact of compromised systems, limits unnecessary connectivity and makes access policies easier to understand. A VLAN is a useful engineering tool. Security comes from what you do with it.