// SECURITY NETWORK DESIGN · HANS STUDY · ONTARIO, CANADA

Security network design

The cameras sit on somebody's network. The door controllers are somebody's endpoints. When the system misbehaves the network is the last thing anyone examines and usually the reason it happened.

Most security systems are specified by people who understand cameras and doors, then installed onto a network designed by nobody in particular. It works on day one because the load is low and the estate is small, and it degrades as both grow. I design the network underneath the systems, and I design it for what security traffic actually does rather than for what office traffic does.

Why security traffic is different

  • It is constant. Video does not have a quiet period. A network sized on average office utilisation is sized wrong.
  • It is sensitive to loss in ways users notice later. Dropped frames do not generate a help desk ticket. They generate a missing incident three months from now.
  • Multicast behaves badly by default. Without querier and snooping configured deliberately, video floods where it should not and starves what should get it.
  • Devices are powered by the network. PoE budget, not port count, is the real constraint, and it is the calculation most often skipped.
  • The endpoints are untrustworthy. Cameras and controllers are embedded devices with long patch cycles sitting in publicly reachable locations.

What the design covers

  • Segmentation. Where security sits relative to corporate, OT and guest, what crosses the boundary, and what enforces it. Separation that survives someone plugging a laptop into a camera drop.
  • VLAN and addressing plan. Structured so it can grow, and documented so the next person can read it.
  • Routing and multicast. Including querier placement and snooping, which is where multi-site video deployments most reliably come apart.
  • Switch selection and sizing. Against PoE draw, uplink capacity and stacking, not against port count alone.
  • Redundancy. Link and device level, sized to what actually needs to survive rather than to everything.
  • QoS. Where it earns its complexity and where it does not.
  • Device hardening posture. What the network does about endpoints that cannot defend themselves.

Platforms

Cisco Catalyst and IOS-XE, HPE Aruba CX and Instant, Juniper EX and Junos, Alcatel-Lucent OmniSwitch, Fortinet, and the usual mixed estates that grew rather than being designed. Vendor selection follows the requirement rather than the other way round, since there is nothing here I earn a margin on.

Who this is for

  • Anyone putting cameras, readers or controllers onto a network at scale.
  • Multi-site owners where each location was built differently and none of them agree.
  • Organisations where IT and physical security have never had to share infrastructure and now do.
  • Projects where the security design is finished and nobody has designed the network yet.

Related

Wireless design and validation →

Where the network is wireless, coverage is measured rather than modelled.

Design it before it is installed

Retrofitting segmentation onto a live security estate is an order of magnitude harder than designing it once. The cheapest version of this work happens before anything is racked.