Use case · Internet breakout
Internet breakout through the firewall you trust
Choose where your offices meet the internet — your existing head-office firewall, or a site you nominate as the exit — and every branch follows, with no separate branch office firewall to buy, configure and keep up to date at each site.
What internet breakout means
Internet breakout is where a site's internet traffic leaves your network. There are two broad choices, and most organisations end up using both:
- Local internet breakout: each office sends its internet traffic straight out of its own line. Fastest for the user, but every office then needs its own firewall, its own rules and its own logs.
- Centralised internet breakout: branches send their internet traffic across the network to one place, usually head office, and it leaves through the firewall there. One set of rules, one place to log, one public address for suppliers to allow-list.
| Local breakout | Centralised breakout | |
|---|---|---|
| Firewalls to run | One per office | One, at the exit |
| Rules and logging | Kept in step across every office | In one place |
| Address the internet sees | Each office's own | The exit's, for every office |
| Speed to cloud services | Shortest path | A detour through the exit |
| If the exit's line fails | Only that office is affected | Needs a second exit to fall back to |
The usual answer is centralised breakout for most traffic, with named exceptions, such as Microsoft 365 and video calls, breaking out locally. Weft does both at once.
Why it goes wrong on ordinary routers
Centralised breakout means pushing a default route — "the internet is that way" — into every branch. On a branch router that is the most dangerous route there is. Get it wrong and the encrypted tunnels follow it too, so the branch sends its own tunnel traffic into the tunnel, cuts itself off, and takes the means to fix it with it. That is why many organisations give up and put a firewall at every branch.
How Weft does internet breakout
- The default route can never capture the tunnels. It lives only inside your offices' own routing table, never in the table that carries the links between sites, so it cannot take down the links that deliver it.
- Hand off to your own firewall. A head-office site sends internet traffic to the firewall on its network — a FortiGate, pfSense, Meraki or anything else that routes and NATs — with each packet's original source address, so the firewall's rules and logs see which office it came from. Routes can be exchanged by OSPF or eBGP if you want.
- Or make a site the exit. Any site can be the internet exit for the others, including one in a cloud region, when there is no firewall to hand off to.
- The nearest exit, with two or more. With a firewall at more than one head office, each branch can use the one with the lowest measured round trip. In our two-continent test, a branch in Virginia went from 0.465 s per request through London to 0.040 s through Virginia, and fell back to London 24 seconds after Virginia's box was stopped.
- Named exceptions break out locally. Microsoft 365, Zoom, Webex, Google Meet and others can leave each branch's own line while everything else goes through the firewall.
- Remote staff can follow it too. A roaming client on full tunnel sends its internet traffic the same way, so it leaves through the same firewall as the offices'.
What your firewall needs to do
Because traffic arrives with the branch's own addresses, the firewall has three jobs, and the console lists them next to the setting:
- Accept traffic from the other offices' networks, not only its own subnet.
- NAT it on the way out, as it already does for head office.
- Route replies for the other offices back through the Weft box on head office's network.
On a Linux firewall in our test that was four lines. Branch internet through your head-office firewall walks through it, including the one step that is easy to miss: in our run the branch was online 79 seconds after it was fixed.
Setting it up, in outline
- Connect your offices as in branch office connectivity.
- On head office's site, under Internet exit, choose Sends it to our firewall and enter the firewall's address. Or choose a site to be the exit itself.
- Nothing to do at the branches: they learn the way out from head office.
- Optionally, choose what breaks out locally at each branch instead.
Questions
Do we still need a firewall at each branch?
Not for internet traffic that leaves centrally: it goes through the head-office firewall's rules. Many organisations keep a simple router at each branch for the line itself.
What if the exit goes down?
With a second exit — another head office's firewall, or a site made an exit — branches fall back to it when the first exit's Weft box goes quiet: in our test, 24 seconds later. Weft watches the boxes; the firewall's own internet line is the firewall's to watch, so a second line there is worth having.
Will suppliers see one address for all our offices?
Yes, for traffic that leaves centrally: the internet sees the exit's address, which is useful when a supplier's system allows only listed addresses.
Does Teams have to go through head office as well?
No, and Microsoft recommends it does not. Choose which Microsoft 365 categories break out locally, and which other call services, and the rest still goes through the firewall.