Port Forwarding and Remote Access
How Port Forwarding Works
A site's router performs NAT: many private inside addresses share one public outside address, and inbound connections have no inside destination unless one is configured. A port forwarding rule (also called a port map or virtual server entry) supplies that mapping:
traffic arriving at public-IP : external-port → send to internal-host : internal-port
Key details:
- TCP vs UDP matters. Forwarding rules are per-protocol. A RoIP product may use TCP (Transmission Control Protocol) for control/login and UDP (User Datagram Protocol) for voice; forwarding only the TCP port produces the classic "logs in fine, no audio." Forward every documented port, with the correct protocol, in the correct direction. See Networking UDP.
- The external and internal port numbers may differ (translation), but keeping them identical avoids confusion unless there's a conflict.
- The internal host must have a static or reserved IP address — a forward pointing at yesterday's DHCP (Dynamic Host Configuration Protocol) address silently breaks.
- The router's own firewall must also permit the forwarded traffic; on many devices the forward rule implies this, on others it's separate — see Firewall Types: Stateful vs Stateless and NAT (Network Address Translation).
Dynamic DNS
Many sites have no static public IP — the ISP (Internet Service Provider) changes it occasionally. Remote clients configured with a bare IP address then break without warning. Dynamic DNS (DDNS) solves this: a small client on the router or a PC notices the public address changed and updates a hostname (for example, one provided by a DDNS service) to match. Remote clients use the hostname instead of the address. Caveats: there is a propagation delay after each change, the DDNS updater must actually be running somewhere on-site, and a business-class static IP is still preferable for critical dispatch links.
Double NAT and Carrier-Grade NAT
Port forwarding assumes the router you configure owns the public address. Two common situations break that assumption:
- Double NAT: two NAT layers in series — for example, the ISP's modem/gateway doing NAT in front of the site's own router. An inbound connection must be forwarded through both layers, configured consistently. Signs: the site router's WAN address is itself a private address (10.x.x.x, 172.16–31.x.x, 192.168.x.x). Remedies: put the ISP device in bridge mode, or configure matching forwards on both devices.
Security: Prefer VPNs Over Raw Exposure
Every forwarded port is a doorway the entire internet can knock on. Internet-wide scanners find newly opened ports within minutes, and radio/dispatch services were rarely designed to withstand hostile traffic; weak or default credentials on an exposed service are a compromise waiting to happen.
The preferred pattern is a VPN: expose only the VPN endpoint (designed for hostile networks, cryptographically authenticated), and let radio traffic flow inside the tunnel with no service directly exposed. If raw forwarding is unavoidable, restrict the rule to known source addresses where the router supports it, keep the exposed service patched, use strong credentials, and forward the absolute minimum.
MTU and Fragmentation Pitfalls over VPN
VPN encapsulation adds header bytes to every packet, shrinking the usable payload below the usual 1500-byte Ethernet MTU (Maximum Transmission Unit). If oversized packets can't be fragmented or the fragmentation signaling is blocked, large packets silently vanish while small ones pass — producing weird, selective failures: pings work but logins hang, small control messages flow but configuration screens or file transfers stall. Voice packets are small and usually unaffected, which makes the pattern more confusing. Remedies: lower the tunnel MTU, enable MSS (Maximum Segment Size) clamping on the router, and test with don't-fragment pings of increasing size to find the real path MTU.
Testing from Outside
Always test remote access from genuinely outside the site network — testing from inside proves almost nothing, and many routers don't support "hairpin" access to their own public address from inside anyway:
- Use a phone on cellular data (Wi-Fi off) or a remote host as the test client.
- Confirm the public IP or DDNS name resolves to what the router currently reports.
- Test the actual application first; if it fails, test the specific ports with a port-checking tool or a simple connect test. Note that UDP is hard to test with generic port checkers (no handshake to observe) — the application itself, or a packet capture at the site while testing, is the reliable check.
- Capture at the internal host if needed: if the test traffic never appears there, the forward/firewall is wrong; if it arrives and isn't answered, the host/application is wrong. See Troubleshooting RoIP and Gateway Connectivity.
Related Pages
- NAT (Network Address Translation)
- Networking VPNs
- Firewall Types: Stateful vs Stateless
- Troubleshooting RoIP and Gateway Connectivity
References
- RFC 6598 — IANA-Reserved IPv4 Prefix for Shared Address Space — defines 100.64.0.0/10 (100.64.0.0 – 100.127.255.255) as Shared Address Space, allocated to "number the interfaces that connect CGN devices to Customer Premises Equipment (CPE)". This is the authority for reading a
100.64.x.xWAN address as CGNAT. (Verified August 2026.) - RFC 1918 — Address Allocation for Private Internets — the 10/8, 172.16/12, and 192.168/16 private ranges used to spot double NAT.
- RFC 3022 — Traditional IP Network Address Translator (Traditional NAT)
- RFC 1191 — Path MTU Discovery — background for the MTU/fragmentation failure mode described above.
- RFC 6888 — Common Requirements for Carrier-Grade NATs (CGNs)
See also: CTI product documentation
This article is part of the CTI Products Knowledge Base. Questions about RadioPro or MCN products? Call +1 (513) 595-5900 or email info@ctiproducts.com.
Informational only. This article is general background and is not engineering, safety, legal or regulatory advice. Verify requirements with the equipment manufacturer and the applicable authority before relying on it. Product names and trademarks belong to their respective owners.