Network Architecture¶
Physical Infrastructure¶
Internet Connectivity¶
- Service: 1Gb/s synchronous fiber
- Router: Ubiquiti EdgeRouter-X (central routing)
Switching Infrastructure¶
Internet (1Gb/s fiber)
│
├── EdgeRouter-X (Central Router)
│
├── Cisco Catalyst WS-C3850-12X48U (SW-AGGR-01 - Primary aggregation switch)
│
├── Cisco Catalyst WS-C3850-12X48U (SW-RACK-01 - Server rack switch)
│ │ 40Gbps uplink (pending installation)
│ │
│ ├── emerald (R720XD)
│ ├── fuji (R720XD)
│ ├── apollo (R720XD + MD1200)
│ └── bishop/castle/domino (R630s)
│
└── Multiple Cisco SG-300 series (House access switches)
Wireless Infrastructure¶
- Access Points: 3x Cisco Catalyst 9120AXI
- Standard: Wi-Fi 6 (802.11ax)
- Controller: Mobility Express firmware (controller-less)
- Coverage: Whole house wireless coverage
- Upgraded: November 2025 (replaced Aironet 3802i)
VLAN Configuration¶
Network Segmentation Strategy¶
All networks use 172.16.x.0/24 addressing scheme:
VLAN 0 - Management Network¶
- Subnet: 172.16.0.0/24
- Purpose: Infrastructure management
- Devices:
- All network switches
- Server iDRAC/IPMI ports
- Network equipment management interfaces
VLAN 90 - Server Network¶
- Subnet: 172.16.90.0/24
- Purpose: Server management and host OS
- Devices:
- Proxmox web interfaces
- unRAID management (apollo)
- Server host OS interfaces
- Administrative access to hypervisors
VLAN 100 - User Network¶
- Subnet: 172.16.100.0/24
- Purpose: Primary user devices
- Devices:
- Family computers and laptops
- Phones and tablets
- Trusted personal devices
VLAN 101 - Guest Network¶
- Subnet: 172.16.101.0/24
- Purpose: Guest device isolation
- Status: Currently unused (guest WiFi configuration pending)
- Planned use: Visitor device access with limited connectivity
VLAN 102 - IoT Network¶
- Subnet: 172.16.102.0/24
- Purpose: Internet of Things device isolation
- Devices:
- Apple TVs
- Home automation devices
- Amazon Echo devices
- Smart home sensors and controllers
- Security: Isolated from main user network
VLAN 103 - Lab Network¶
- Subnet: 172.16.103.0/24
- Purpose: Virtual machine and lab workloads
- Devices:
- All Proxmox VMs
- Kubernetes cluster nodes (primary network interface)
- Development and testing workloads
- Lab experiments and projects
VLAN 104 - Data-Sync/Storage Network¶
- Subnet: 172.16.104.0/24
- Purpose: Dedicated 10GbE storage traffic
- Status: Deployed and operational (December 2025)
- MTU: 9000 (jumbo frames enabled)
- Devices:
- Apollo: 172.16.104.30 (Dell Y40PH 10GbE NIC)
- Emerald: 172.16.104.34 (Built-in Broadcom BCM57810 10GbE)
- Fuji: 172.16.104.35 (Dell Y40PH + SFP-10G-T-X transceiver)
- Production cluster VMs (secondary network interface eth1)
- Switch: SW-RACK-01 (Cisco 3850) ports Te1/0/37-42
- Performance: Sub-millisecond latency, 10Gbps link speed
- Use cases:
- Ceph distributed storage cluster traffic
- High-bandwidth VM storage operations
- NFS storage traffic from apollo
- Database replication and backups
VLAN 110 - LoadBalancer Network¶
- Subnet: 172.16.110.0/24
- Purpose: MetalLB IP pool for Kubernetes services
- BGP peering: Advertised to EdgeRouter-X via BGP
- Services:
- Traefik ingress controller: 172.16.110.0
- User-facing Kubernetes services
Kubernetes Network Integration¶
Cluster Networking Strategy¶
Management Access¶
- Proxmox hosts: VLAN 90 (172.16.90.x)
- Kubernetes API: Accessible from VLAN 90 and 100
- Administrative access: VLAN 90 for management operations
VM Network Placement¶
Minimal Cluster VMs (VLAN 103 + VLAN 104):
├── minimal-cp-01: 172.16.103.10 (eth0) / 172.16.104.10 (eth1)
├── minimal-wk-01: 172.16.103.11 (eth0) / 172.16.104.11 (eth1)
└── minimal-wk-02: 172.16.103.12 (eth0) / 172.16.104.12 (eth1)
Production Cluster VMs (planned - VLAN 103 + VLAN 104):
├── prod-cp-01: 172.16.103.110 (eth0) / 172.16.104.110 (eth1)
├── prod-cp-02: 172.16.103.111 (eth0) / 172.16.104.111 (eth1)
├── prod-cp-03: 172.16.103.112 (eth0) / 172.16.104.112 (eth1)
├── prod-worker-01: 172.16.103.120 (eth0) / 172.16.104.120 (eth1)
├── prod-worker-02: 172.16.103.121 (eth0) / 172.16.104.121 (eth1)
├── prod-worker-03: 172.16.103.122 (eth0) / 172.16.104.122 (eth1)
├── prod-worker-04: 172.16.103.123 (eth0) / 172.16.104.123 (eth1)
├── prod-storage-01: 172.16.103.170 (eth0) / 172.16.104.170 (eth1)
└── prod-storage-02: 172.16.103.171 (eth0) / 172.16.104.171 (eth1)
Note: All VMs use dual network interfaces:
- eth0 (VLAN 103): Kubernetes API, pod communication, service traffic
- eth1 (VLAN 104): Ceph storage traffic, high-bandwidth I/O
Service Exposure Strategy¶
- Internal services: Stay within VLAN 103
- User-facing services: Expose to VLAN 110 via load balancer
- Management interfaces: Accessible from VLAN 90
Load Balancing and Ingress¶
- MetalLB: Kubernetes load balancer for service exposure
- IP pools: Carved from VLAN 103 subnet
- Ingress controller: Nginx or Traefik for HTTP/HTTPS routing
Security Considerations¶
Network Isolation¶
- Management isolation: VLAN 0 separated from user traffic
- IoT containment: VLAN 102 prevents lateral movement
- Guest isolation: VLAN 101 limited connectivity when enabled
- Lab sandboxing: VLAN 103 for experimental workloads
Firewall Rules (EdgeRouter-X)¶
VLAN 100 → VLAN 103: Allow HTTP/HTTPS to services
VLAN 90 → All: Management access
VLAN 0 → All: Infrastructure management
VLAN 102 → Internet: Allow, block inter-VLAN
VLAN 103 → Internet: Allow outbound
Access Control¶
- Administrative access: Requires VLAN 90 connectivity
- User services: Exposed selectively to VLAN 100
- Inter-VLAN routing: Controlled by EdgeRouter-X policies
Storage Network Architecture¶
Deployed: 10GbE Storage Network (VLAN 104)¶
Status: Operational since December 2025
Physical Infrastructure¶
- Network: VLAN 104 (172.16.104.0/24)
- Switch: SW-RACK-01 (Cisco Catalyst 3850)
- MTU: 9000 (jumbo frames)
- Link speed: 10Gbps
- Topology: Layer 2 switched network
Connected Servers (3 of 6)¶
| Server | Interface | IP Address | Switch Port | Hardware |
|---|---|---|---|---|
| Apollo | eth4 | 172.16.104.30 | Te1/0/37 | Dell Y40PH (Broadcom bnx2x) |
| Emerald | enp68s0f0/vmbr2 | 172.16.104.34 | Te1/0/41 | Built-in Broadcom BCM57810 10GbE |
| Fuji | enp68s0f0/vmbr2 | 172.16.104.35 | Te1/0/42 | Dell Y40PH + SFP-10G-T-X transceiver |
Pre-configured Ports (Ready for expansion)¶
- Te1/0/38: bishop-data-sync-10g (VLAN 104, MTU 9000)
- Te1/0/39: castle-data-sync-10g (VLAN 104, MTU 9000)
- Te1/0/40: domino-data-sync-10g (VLAN 104, MTU 9000)
Performance Validation¶
- Link speed: 10Gbps confirmed on all interfaces
- Jumbo frames: 8972-byte ICMP payload working
- Latency: Sub-millisecond (~0.3-0.4ms between servers)
- Persistence: All configurations survive reboots
Use Cases¶
- Ceph distributed storage: High-bandwidth replication between OSDs
- VM storage operations: Fast block device I/O
- NFS traffic: High-performance NFS from apollo
- Database replication: Low-latency, high-bandwidth DB clustering
- Backup operations: Rapid large data transfers
- Container registry: Fast image distribution
Dual-Network VM Architecture¶
Kubernetes cluster VMs use two network interfaces:
- eth0 (VLAN 103): Control plane, API, pod networking
- eth1 (VLAN 104): Storage traffic (Ceph, NFS, backups)
Benefits:
- Isolates storage traffic from cluster management
- Prevents storage operations from saturating cluster network
- Enables 10x bandwidth increase for storage workloads
- Simplifies network troubleshooting and monitoring
Future Expansion¶
Phase 1: Bishop, Castle, Domino (Planned)¶
- Connect remaining 3 servers to 10GbE network
- Switch ports already configured and ready
- Requires 10GbE NICs or SFP+ transceivers
Phase 2: 40Gbps Inter-Switch Uplink (Planned)¶
- Hardware: C3850-NM-2-40G modules (installed)
- Cable: QSFP+ DAC between SW-AGGR-01 and SW-RACK-01
- Benefit: 40Gbps aggregate bandwidth for 6x 10GbE servers
Monitoring and Management¶
Network Monitoring¶
- SNMP: Available on all Cisco equipment
- Bandwidth monitoring: EdgeRouter-X and switch statistics
- Connectivity monitoring: Ping and service checks
Integration with Kubernetes¶
- Prometheus: Monitor cluster network performance
- Grafana: Dashboard for network metrics
- Alert manager: Network connectivity alerts
Hybrid Cloud Networking¶
Oracle Cloud Integration¶
- VPN connectivity: Site-to-site VPN for hybrid workloads
- Network segmentation: Separate subnets for cloud workloads
- Security: Encrypted tunnels for sensitive data transfer
DNS and Service Discovery¶
- Internal DNS: Resolve internal services
- External DNS: Public service exposure
- Service mesh: Future consideration for microservices
Performance Characteristics¶
Bandwidth Allocation¶
- Internet: 1Gb/s shared across all VLANs
- VLAN 103 (Cluster): 1Gb/s per VM
- VLAN 104 (Storage): 10Gb/s per server/VM
- Switching capacity: 176Gb/s (Cisco Catalyst 3850)
- Inter-switch capacity: 40Gb/s QSFP+ uplink (pending installation)
Latency Expectations¶
- Same VLAN: <1ms
- Inter-VLAN: <2ms (routing overhead)
- Internet: 10-20ms (fiber connection)