Skip to main content

Network Architecture

DTC's network architecture is built on firewall-centric design with complete network segmentation. The firewall handles DNS, DHCP, routing, IPS, and traffic visibility. Every device starts isolated and communicates only through explicitly permitted paths.

Core Principles

  • Firewall-centric: The firewall is the sole DNS server, sole DHCP server, sole gateway, and sole point of internet egress for every device on the network
  • Segmentation by device role: Separate VLANs for infrastructure, networking, production, DMZ, cameras, voice, and guests
  • DHCP reservations over static IPs: Only DNS servers, DHCP servers (the firewall), and gateways get static IPs. Everything else uses DHCP reservations.
  • Automatic firmware updates: Managed device firmware auto-updates daily outside business hours. Gateway/firewall firmware auto-updates weekly outside business hours. No manual maintenance windows for routine firmware.
  • IPS on every VLAN: Intrusion Prevention System enabled on all VLANs, all detection categories, notify and block mode

VLAN & IP Addressing Scheme

All sites use a consistent VLAN numbering and subnet scheme. The third octet of each subnet mirrors the VLAN ID, so knowing one tells you the other. The location number maps to the second octet (location 1 = 10.1.x.x, location 2 = 10.2.x.x, etc.). Gateways are always the .1 of each subnet.

VLANVLAN IDSubnet (within location /16)Example (location 1 = 10.1.0.0/16)
NETWORK1x.x.0.0/2410.1.0.0/24
INFRA10x.x.10.0/2410.1.10.0/24
CORP20x.x.20.0/2410.1.20.0/24
BLDG2-CORP21x.x.21.0/2410.1.21.0/24
BLDG3-CORP22x.x.22.0/2410.1.22.0/24
DMZ-PUBLIC30x.x.30.0/2410.1.30.0/24
DMZ-INTERNAL31x.x.31.0/2410.1.31.0/24
CAMERAS40x.x.40.0/2410.1.40.0/24
VOIP50x.x.50.0/2410.1.50.0/24
GUEST99192.168.240.0/20192.168.240.0/20

Rules:

  • VLAN IDs are fixed and consistent across all sites — do not deviate.
  • The third octet mirrors the VLAN ID for all 10.x.x.x VLANs (e.g., VLAN 40 = x.x.40.0/24). GUEST is the intentional exception.
  • Location number = second octet. Location 1 is 10.1.0.0/16, location 2 is 10.2.0.0/16, and so on.
  • Gateway is always .1 of each subnet (e.g., 10.1.20.1 for CORP at location 1).
  • GUEST is fixed at 192.168.240.0/20 across all sites. It is fully isolated and has no internal routing, so a consistent fixed range is safe and simplifies configuration.
  • Multi-building CORP VLANs increment from 20 (21, 22...). Single-building sites use VLAN 20 / CORP only.

VLAN Architecture

Site-Wide VLANs

These VLANs are shared across the entire site regardless of building count.

VLANPurposeDevicesZone
INFRAStorage and compute infrastructureHypervisors, NAS, BDR/backup appliances, container hosts, SANInternal, Isolated
NETWORKNetwork infrastructure managementSwitches, routers, firewalls (management), APsInternal, Isolated
DMZ-PUBLICInternet-facing servicesReverse proxies, web servers, services accessible from outsideInternal, Not Isolated
DMZ-INTERNALInternal-only segmented servicesInternal apps that need segmentation from CORP but aren't internet-facingInternal, Not Isolated
CAMERASSurveillance infrastructureNVRs, IP cameras, video management systemsInternal, Isolated
VOIPVoice infrastructureVoIP phones, call servers, SBCsInternal, Isolated
GUESTUntrusted accessVisitor WiFi, untrusted devicesHotspot, Isolated

Per-Building VLANs

CORP VLANs are created per building/location to segment production endpoints by physical area.

VLANPurposeDevices
BLDG1-CORPBuilding 1 production endpointsWorkstations, printers, internal servers, DCs
BLDG2-CORPBuilding 2 production endpointsWorkstations, printers, internal servers
ANNEX-CORPAnnex production endpointsWorkstations, printers, internal servers

The naming pattern is {building}-CORP. Single-building sites just use CORP. Multi-building sites get one CORP VLAN per building.

Key Placement Rules

  • INFRA: Hypervisor hosts go here at the physical port level. VMs get VLAN-tagged at the VM level to whatever VLAN they belong on (CORP for DCs/file servers, INFRA for storage VMs, etc.)
  • NETWORK: Applied to the device itself, not the switch port
  • CORP (per-building): Domain controllers and servers that need direct endpoint access live here (AD uses too many dynamic ports for cross-VLAN pinholing). Each building's endpoints are isolated from other buildings at L2.
  • CAMERAS: NVRs and cameras isolated from production traffic. Firewall rules allow NVR access from CORP as needed for viewing.
  • VOIP: Phones isolated from production traffic. QoS can be applied per-VLAN for voice prioritization.
  • GUEST: L3 isolation + device isolation. No access to internal resources. Zone must be Hotspot.
  • DMZ-PUBLIC/DMZ-INTERNAL: Open by default. L3 network isolation using SPI firewall rules controls what is NOT permitted. Each DMZ VLAN has specific deny rules for traffic that shouldn't reach it.

IP Address Class Selection

IP class is chosen based on business size and location requirements. Each location gets a contiguous block, subdivided into per-VLAN subnets.

RangeUse CasePer-Location AllocationSubnet Range
172.16.0.0/12Docker containers and virtual machines/20 per location/20 to /24 per VLAN
10.0.0.0/8Small and large business / locations/16 per location/20 to /24 per VLAN

Rules:

  • These IP addressing standards apply to new deployments and new locations only. Do not re-scheme existing client networks unless required by a specific change (e.g., integrating Active Directory across locations, site-to-site VPN conflicts) or the client explicitly requests it. The disruption of re-addressing an existing production network is almost never worth it.
  • Default is 10.0.0.0/8. All DTC clients — small and large — use this range. Start here for all new deployments.
  • Each location gets a /16 allocation within the 10.0.0.0/8 space. Assign sequential /24s within the block for each VLAN (e.g., 10.0.0.0/24, 10.0.1.0/24, 10.0.2.0/24...). Next location starts at the next /16 boundary (e.g., 10.1.0.0/16, 10.2.0.0/16...).
  • 172.16.0.0/12 is reserved for Docker/container hosts and virtual machine overlay networks... never for physical endpoints.
  • VLAN IDs just need to be unique. Specific numbers don't matter.
  • Within a location's allocation, assign sequential subnets per VLAN.

Firewall and Security

Intrusion Prevention (IPS):

  • Enabled on ALL VLANs... no exceptions
  • All detection categories enabled (botnets, malware, exploits, P2P/dark web, attacks, protocol vulnerabilities)
  • Detection mode: Notify and Block

DNS Security:

  • Upstream DNS: Cloudflare (1.1.1.1) or DNSFilter depending on the client
  • Do NOT use Google (8.8.8.8), Quad9 (9.9.9.9), ISP DNS, or any other upstream provider. Cloudflare or DNSFilter only.
  • Gateway DNS logging enabled for visibility

Traffic Logging:

  • NetFlow (IPFIX): Off unless a NetFlow collector/endpoint is configured to receive the data. Don't enable logging with nowhere to send it.
  • Flow logging: All Traffic (required for visibility and troubleshooting)
  • Gateway DNS logging enabled
  • Activity logging: Forwarded to SIEM (e.g., Blumira provided by DTC) if the client has one, otherwise stored internally on the firewall

Guest Isolation:

  • GUEST VLAN zone: Hotspot (not Internal)
  • L3 Network Isolation: GUEST isolated from all networks
  • Device Isolation: Enabled on GUEST
  • No access to internal resources

Inter-VLAN Access:

  • Default posture: Allow All (firewall rules restrict as needed)
  • BDR/NAS on INFRA need an internal-to-internal allow rule for backup traffic from CORP
  • Production VMs on hypervisors (INFRA port) get VLAN tagged at the VM level to CORP
  • CAMERAS VLAN: Allow NVR viewing access from CORP via firewall rule
  • VOIP VLAN: Allow call control traffic as needed

Switch Standards

  • Spanning Tree Protocol: RSTP
  • Rogue DHCP Server Detection: Enabled
  • Jumbo Frames: Disabled
  • 802.1X: Disabled (unless specific client requirement)

Port assignments:

Port TypeConnectionNative VLANAllowed VLANs
AccessWorkstationCORPBlock all others
AccessPrinterCORPBlock all others
AccessPhoneVOIPBlock all others
AccessIP CameraCAMERASBlock all others
AccessNVRCAMERASBlock all others
TrunkHypervisorINFRAAllow all
TrunkNAS/BDRINFRAAllow all
TrunkSwitch to SwitchNoneAllow all
TrunkSwitch to APNoneAllow all

Access Ports: Native VLAN set to the device's assigned VLAN. Block all other VLANs.

Trunk Ports: Allow all VLANs. Switch-to-switch and switch-to-AP trunks have no native VLAN... the management interface on each device uses its assigned VLAN (NETWORK) or VLAN 1. Hypervisor and NAS/BDR trunks use INFRA as the native VLAN because that's the host's management VLAN.


Wireless Standards

  • All APs must be hardwired... no wireless meshing
  • Band steering disabled (causes issues with Apple devices)
  • All bands enabled: 2.4 GHz, 5 GHz, 6 GHz (if available)
  • Extended 5 GHz Spectrum (DFS) enabled
  • Channel AI enabled for automatic optimization

Channel Widths:

BandUncrowdedCrowded/Noisy
2.4 GHz40 MHz20 MHz
5 GHz160 MHz80 MHz
6 GHz320 MHz160 MHz

When in doubt, use crowded settings. Most client sites are in shared office buildings.

Guest WiFi: All standard WiFi settings plus multicast/broadcast blocker, multicast-to-unicast, client device isolation, BSS transition. Security: WPA2, PMF disabled.


WAN and Internet

  • WAN mode: Failover Only (not load balancing)
  • Automatic speed tests: Daily outside business hours
  • Flow Control: Disabled (enable only if bandwidth constrained and ISP speeds test well)
  • Encrypted DNS: Off (we control DNS at the firewall level)
  • Region blocking: Disabled

Firmware and Software Updates

All managed network equipment should be configured for automatic firmware updates outside of business hours. This ensures security patches are applied promptly without requiring manual maintenance windows.

Principles:

  • Managed devices (switches, APs, etc.): Auto-update firmware daily outside business hours
  • Gateways/firewalls: Auto-update firmware weekly outside business hours (less frequent due to higher impact of reboots)
  • All update schedules must be outside client production hours and staggered so they do not fire at the same time
  • Vendor-specific schedules and settings are documented in the vendor's book (e.g., Ubiquiti book for UniFi devices)

Physical Infrastructure

  • Managed switches required... no unmanaged switches in production
  • All APs hardwired with structured cabling
  • Cable management and labeling standards per site
  • UPS protection for network equipment where possible