A site-to-site VPN connects the local networks of two or more locations through encrypted tunnels across the internet. The tunnels run between gateways at each site, so users in a branch reach the ERP, file server or cameras at head office as if they were on the same network, without running any client. It is the most common way for a multi-branch SMB to operate as one network without paying for leased lines.

Site-to-site vs remote access

Site-to-siteRemote access
What connectsNetwork to networkDevice to network
ClientNone; gateways build the tunnelVPN client on the user's device
ConnectionPersistent, always onWhen the user needs it
AuthenticationBetween gateways, with certificates or a strong shared keyPer user, with MFA
Typical useBranches, shops, warehouses, hotelsHome working, field teams

Most SMBs need both: branches join head office over site-to-site and remote staff over remote access. Managing both on the same gateway with the same access policy keeps things simple.

Choosing a topology

  • Hub-and-spoke: every branch connects to head office. Simple to build and run; branch-to-branch traffic passes through the hub.
  • Full mesh: every site connects directly to every other. n sites need n(n−1)/2 tunnels, so 10 sites means 45 tunnels.
  • Hybrid: hub-and-spoke as the base, with direct tunnels added between site pairs that exchange a lot of traffic.

For SMBs whose applications run at head office or in the cloud, hub-and-spoke is usually enough. Whether traffic to cloud applications such as email and CRM detours through head office is settled by the split tunnel decision below.

Design steps

1. A non-overlapping IP plan

Many branch routers default to 192.168.1.0/24. If two sites use the same subnet, routes collide even when the tunnel is up. Give every site a unique subnet (e.g. 10.10.site.0/24) and plan it before connecting the first branch. If overlap can't be avoided, NAT over the tunnel is possible but makes troubleshooting harder.

2. Routing

With a handful of sites, static routes are enough. As the site count grows, managing which subnet is reached through which tunnel centrally becomes important.

3. Split tunnel or full tunnel?

  • Full tunnel: all of the branch's internet traffic also exits through head office. Security inspection happens in one place, but head office's internet link and latency take the load.
  • Split tunnel: only business traffic enters the tunnel and internet traffic breaks out locally. More efficient, but the branch exit needs security inspection too.

4. Encryption

IPsec with IKEv2 has long been the most common standard for site-to-site, and newer lightweight protocols are also in use. Whichever protocol you choose, rely on current encryption algorithms, certificates or a strong shared key, and regular rekeying.

5. MTU

Tunnel encapsulation adds a header to every packet. Large packets get fragmented or dropped, and some applications seem to "connect but crawl". Configure MSS clamping on the tunnel interface.

6. DNS

For branch users to reach head-office servers by name (e.g. erp.company.local), queries for internal domains must be forwarded over the tunnel to the DNS server at head office. Skip this step and the tunnel works, but users have to connect by IP address and the problem gets reported as "the VPN is down".

7. Redundancy

Head office is where every branch connects, so two separate internet links with failover matter. Tunnels should re-establish automatically after a drop, and their state should be monitored. A mobile data line as a second link is a low-cost backup for critical branches.

Security: a tunnel is not trust

An encrypted tunnel protects traffic in transit, but it says nothing about whether the devices at either end are safe. A compromised PC in one branch can reach head-office servers over an unrestricted tunnel.

  • Least privilege: shop POS terminals should reach only the ERP at head office, not the whole network. Define rules per user and group in a centralised policy.
  • MFA: require a second verification step on top of the password for remote access users.
  • Inspect tunnel traffic: traffic between sites should pass IPS and application control too.
  • Revoke access quickly: when someone leaves, one action should end their access on every device.
  • Keep records: who connected, when and from where; session history matters for troubleshooting and audits.

Rollout order

  1. List the sites and the applications each one needs.
  2. Draw up a non-overlapping IP plan.
  3. Decide on the topology and on split or full tunnelling.
  4. Write the access policy: who reaches which system?
  5. Test MTU, application performance and failover at a pilot branch.
  6. Connect the remaining branches in stages and put tunnel state under monitoring.

With CoreGate VPN Gateway

CoreGate VPN Gateway provides persistent, encrypted site-to-site tunnels that re-establish automatically after a drop. Branches can use a CoreGate deployment or a compatible existing router. Remote staff connect with a managed VPN client on desktop and mobile devices, and per-user and per-group access policy, MFA and connection reports are managed from one console. Tunnel traffic passes IPS and application control on the same platform through CoreGate Security Gateway. For guest networks in hotels, see our hotel guest Wi-Fi guide.