AWS Direct Connect vs Site-to-Site VPN Failover

Direct Connect vs VPN isn't really either/or for production — it's a primary-plus-failover pattern. Here's how to design it, and when either/or is right.

Frequently Asked Questions

Can a single Site-to-Site VPN tunnel fully back up a 10 Gbps Direct Connect circuit?

No. A single VPN tunnel caps at 1.25 Gbps and AWS explicitly does not recommend it as backup for Direct Connect circuits faster than 1 Gbps — failover would collapse throughput to a fraction of normal capacity.

How do you get more than 1.25 Gbps of VPN failover bandwidth on AWS?

Terminate multiple VPN tunnels on a Transit Gateway instead of a Virtual Private Gateway and enable Equal Cost Multi-Path (ECMP) routing across them — this can scale aggregate backup throughput up to roughly 50 Gbps.

Why would traffic route over VPN even while Direct Connect is healthy?

If the VPN path advertises a more specific BGP prefix than Direct Connect, AWS will prefer the more specific route regardless of which connection is "up" — this fails silently, with no alarm, because nothing is technically down, just misrouted.

Is it enough to confirm the VPN tunnel shows "up" in the AWS console to trust your failover?

No. A tunnel showing "up" only confirms the tunnel exists — it doesn't confirm traffic actually reroutes correctly. Periodically simulate a real Direct Connect outage and verify traffic fails over and fails back cleanly.

Does Direct Connect or VPN provide better encryption by default?

Site-to-Site VPN encrypts traffic over IPSec tunnels by default; Direct Connect is a private physical circuit that bypasses the public internet but is not encrypted by default, so workloads requiring encryption in transit over Direct Connect typically layer MACsec or an application-level TLS/VPN on top.

Discussion0