How auto MPLS traffic balancing works
The short story is that your redundant backhauls are probably carrying traffic badly, with backup links staying idle and traffic being split evenly across uneven-capacity links. We've created a way to fix that in software, getting more capacity out of the network you already have.
Your routing protocol doesn't know what your links can carry
Wireless backhaul capacity isn't a fixed number, it's a moving target. A licensed or unlicensed link renegotiates its modulation (MCS) constantly with rain fade, interference, and alignment drift. A link installed as "700 Mbps" is a 700 Mbps link on a clear day and a 180 Mbps link during a storm cell, and it reports both states as "up."
Every traditional way of handling redundant paths assumes capacity is static. Here's how the three approaches handle the same ring with one degraded link:
RSTP Bridging
Spanning tree exists to prevent loops, so it solves redundancy by blocking it. Your backup path carries zero traffic until the primary fails outright, and a degraded-but-up link never triggers failover at all. The second path you paid to build carries nothing for most of its life.
Backup link carries nothingOSPF + ECMP Routing
Better, since both paths carry traffic. But "equal-cost" is decided by static metrics you typed in. When a link's modulation collapses in weather, its OSPF cost doesn't move, so the protocol keeps sending half the flows into a link that can only carry a fraction of them.
Blind to live capacitySwift Fox RSVP-TE (MPLS)
Swift Fox measures what each link can actually carry right now and solves for balance: traffic is steered through RSVP-TE tunnels — including longer paths a shortest-path protocol would never consider — until every bottleneck sits at the same utilization percentage. The degraded link still works at its real capacity; nothing is the next hotspot.
Every path, equally loadedWhere the capacity numbers come from
Live modulation data, every vendor
Swift Fox polls signal and MCS from your radios — Cambium, Ubiquiti, Mimosa, MikroTik, and every other major vendor — and converts modulation state to usable capacity continuously. When a storm cell knocks a link from MCS 12 to MCS 4, the capacity model updates in seconds, before the congestion it would have caused ever forms.
Automated TCP throughput tests
Vendor-reported capacity can be wrong in ways modulation data never reveals — a 100mbit negotiation on a gigabit port, hidden interference, a failing cable, a misconfigured shaper somewhere in the path. Swift Fox runs automated TCP tests between your MikroTik routers to measure what the path really delivers end to end. These two inputs cover each other's blind spots.
Balancing uses the whole ring, including the reverse direction
A ring has capacity in both directions, and default routing uses only one of them. With balancing on, downstream traffic rides both arms of the ring at once, the eastern arm carrying it “backwards” relative to the default route direction. The paths are equalized all the time, storm or no storm, rather than only during a failure. Flip the toggle to watch the same storm handled both ways:
Plain OSPF: traffic always takes the shortest path, so the eastern arm sits mostly idle — and when the storm degrades WEST–NORTH, its cost never changes, so packets drop while spare capacity sits unreachable. Click Swift Fox balancing above to see the same storm handled with TE tunnels.
Under the hood: Swift Fox uses RSVP-TE to maintain tunnels via MPLS across the topology and allocates traffic to them as measured capacity changes. On MikroTik, that's native in RouterOS v6; on v7 (which dropped TE tunnel routing) we deliver it as a container that runs inside the router itself; and the Pathbox appliance (early access) does it inline for non-MikroTik cores. If you're already running OSPF, enabling MPLS alongside it takes seconds, and the two coexist — TE handles the steering, OSPF remains your fallback.
What this means in dollars
Using capacity you already paid for
Deferred upgrades. When the "full" link toward a tower is really only full because the other direction of the ring sits idle, the upgrade you're pricing out isn't needed yet. Raising usable ring capacity by even 20–30% can often pushes a five-figure backhaul build a year down the road.
Storm days. When a storm knocks a link down to a quarter of its capacity, traffic rebalances onto healthy paths before customers notice. Instead of an outage with angry calls and a midnight truck roll, it shows up as a line item in tomorrow's report.
No new hardware. It runs on the MikroTik core you already own, and if you don't run MikroTik, the Pathbox appliance supports any RSVP-TE capable MPLS network.
See it rebalance your topology, live
Bring a diagram of your ring to a 20-minute demo and we'll walk through exactly what Swift Fox would do with it. You'll be talking to the person who built the feature.
Book a 20-minute demoPrefer email? [email protected]
Next deep dive: how Package-Fair Oversubscription works →