I worked on this multi-site redesign and I am writing about what sits behind the final diagram. Three extra VLANs and a firewall do not magically fix a network. The real subject is what must communicate, what must stay apart and how we prove it without knocking production over.
The starting point
The client operates several production sites linked to an external datacenter over an MPLS network. My scope mainly covers the physical-security LANs: cameras, access control, detection systems, and related equipment.
Two problems overlap. Within each site, too many devices share the same segment. Across the MPLS network, external contractors have a much wider path than they need to perform their work.
- Connectiona contractor joins the network
- Reachthe MPLS exposes several sites
- Movementunnecessary networks become reachable
- Impacta local incident can change scale
What we actually need to fix
- Isolate the sites from one another. A camera or access controller at one site has no reason to initiate traffic towards every other site.
- Constrain contractor access. Each contractor should reach only the required equipment, for the required period, following least privilege.
- Relocate DHCP, DNS, and Active Directory. Two pilot sites must take over services currently hosted in the external datacenter without creating a new single point of failure.
- Redesign addressing and segmentation. The legacy safety LAN is short on space and mixes devices with different roles and trust levels.
Internet egress remains centralised in the external datacenter. That constraint belongs in the design: a useful proposal improves security without pretending that everything can be moved at once.
How I split the subject
Each module attacks a concrete question. What is reachable today? What still needs to work tomorrow? Where should the control sit? Most importantly, how do we prove we did not break anything?