Network design and planning
A network that scales starts with a documented design — topology, IP addressing, VLAN segmentation, redundancy and growth headroom. Segment by function (servers, clients, guests, IoT, management) with VLANs, and route between segments through a firewall or L3 switch with ACLs. Plan redundancy at every layer — dual uplinks, stacked switches, HA firewall pair — and document the failover behaviour so it is not a surprise during an incident.
- Document topology, addressing, VLANs and redundancy
- Segment by function: servers, clients, guests, IoT, management
- Plan dual uplinks and HA firewall pairs
- Leave growth headroom in IP addressing and port capacity
Firewall and VPN configuration
The firewall is the primary security control. Default-deny, allow least privilege, and log denied traffic for review. Use object groups for addresses and services so rules stay readable. For remote access, prefer a modern VPN (WireGuard, IKEv2) or a Zero Trust Network Access layer over legacy SSL VPN clients. Site-to-site VPNs connect branches; use redundant tunnels with dynamic routing (BGP/OSPF) where possible so failover is automatic.
| VPN type | Use case | Notes |
|---|---|---|
| WireGuard | Site-to-site, modern clients | Fast, simple, modern crypto |
| IKEv2 / IPSec | Road warriors, site-to-site | Native on most OS, resilient |
| SSL VPN (legacy) | Legacy remote access | High CVE history — migrate off |
| ZTNA | App-specific access | Identity + device posture, no network access |
Network security best practices
Layer security: segmentation, firewall rules, NAC for endpoint admission, and monitoring. Keep management interfaces on a dedicated VLAN reachable only via a jump host. Disable unused switch ports, enable port security, and use DHCP snooping + dynamic ARP inspection to prevent L2 attacks. Patch network devices on a cadence — a vulnerable firewall is worse than no firewall.
- Default-deny firewall; log denied traffic
- Management VLAN + jump host, no direct internet exposure
- DHCP snooping, dynamic ARP inspection, port security
- Patch network devices on a cadence
Network monitoring and optimization
Monitor with SNMP/flow telemetry into a tool like PRTG, Zabbix, LibreNMS or SolarWinds. Track bandwidth, errors, discards, latency and uptime per interface. Set thresholds with headroom and alert on trends. Optimise by identifying top talkers (NetFlow/sFlow), QoS-marking real-time traffic (voice, video), and right-sizing uplinks before they saturate. Baseline normal behaviour so anomalies are obvious.
// Switch QoS — trust DSCP from voice VLAN, mark best-effort
class-map match-any VOICE match dscp ef
policy-map QOS class VOICE priority percent 30
interface range gi1/0-24 service-policy input QOS
// Monitor: SNMP ifInErrors, ifOutDiscards, ifHCInOctets per interfaceBandwidth management
Bandwidth is finite and expensive. Shape or police non-business traffic, prioritise business-critical apps with QoS, and use caching (web proxy) where appropriate. For multi-site organisations, consider SD-WAN to bond uplinks and route by application. Monitor per-site consumption and forecast capacity — running at 90% utilisation is a pending outage.
- QoS-mark and prioritise real-time and business-critical traffic
- Shape or police non-business traffic
- Consider SD-WAN to bond and steer uplinks by app
- Forecast capacity — 90% utilisation is a pending outage
Network troubleshooting
A disciplined troubleshooting method saves hours. Start at layer 1 (cable, link), then layer 2 (VLAN, MAC, spanning tree), layer 3 (routing, ARP), then the application. Use ping/traceroute for reachability, arp/ip neigh for L2/L3 mapping, and packet capture (tcpdump/Wireshark) for the hard cases. Document the resolution — a recurring issue points to a design problem, not a series of incidents.
| Symptom | Likely layer | First check |
|---|---|---|
| No link | L1 | Cable, port, SFP, interface status |
| No reachability within VLAN | L2 | VLAN membership, spanning tree, MAC table |
| No reachability across VLANs | L3 | Routing, ARP, gateway, ACLs |
| Slow / intermittent | L2–L4 | Errors, discards, duplex mismatch, MTU |
| App fails but network OK | L7 | DNS, cert, firewall app rule, service health |
Lessons learned
- Document the design — undocumented networks fail in surprising ways
- Segment by function and default-deny between segments
- Migrate off legacy SSL VPN to WireGuard/IKEv2 or ZTNA
- Troubleshoot bottom-up: L1 → L2 → L3 → app
Written by
Jamshed Alam
Global MSP & IT Infrastructure Consultant with 20 years across Exchange, Microsoft 365, Azure, security and automation. Available for project and contract work worldwide.