Cloudflare has added flow-based monitoring for Magic Transit, aimed at customers who run the service in on-demand mode. The feature lets Cloudflare analyze traffic flows from a customer's data centers, detect volumetric DDoS activity and notify the customer, so mitigation can begin before an attack takes a business toll.

The gap in on-demand deployments

Magic Transit protects an entire network by putting Cloudflare's network in front of it, with DDoS protection, traffic acceleration and other network functions served from every Cloudflare data center. Cloudflare advertises customer IP prefixes to the Internet over BGP to pull traffic in for scrubbing. The service runs at 59 Tbps of global capacity across 200+ cities, with mitigation at the edge in under 3 seconds.

Always-on customers have their prefixes advertised and attacks mitigated continuously. On-demand customers activate advertisement only when an attack is underway. That leaves a window: while traffic is not routed through Cloudflare, an attack can do damage before anyone notices and turns on Magic Transit.

How flow-based monitoring closes it

With flow-based monitoring, routers are configured to send NetFlow (with sFlow coming soon) continuously to Cloudflare. Cloudflare ingests the flow data and analyzes it for volumetric DDoS attacks.

Flow-based monitoring for Magic Transit Embedded Image - xyqVlv

Attack detection triggers an automatic notification through email, webhook and/or PagerDuty, carrying details about the attack.

Flow-based monitoring for Magic Transit Embedded Image - lT8MKn

Activation of IP advertisement can be handled manually through the Cloudflare dashboard or API, or automatically to minimize time to mitigation. Once Magic Transit is active and traffic flows through Cloudflare, only clean traffic returns to the customer network over GRE tunnels.

Flow-based monitoring for Magic Transit Embedded Image - m3zdWD

The setup is a one-time configuration, after which Cloudflare either alerts the team under attack or starts protecting the network automatically.

Customer walkthrough: Acme Corp

A hypothetical deployment shows the workflow. After a large ransom DDoS attack caused downtime for both external-facing and internal applications, Acme Corp's network team adopts on-demand Magic Transit. The team authorizes Cloudflare to advertise its IP space in the event of an attack, builds Anycast GRE tunnels to receive clean traffic, and configures routers in each data center to send NetFlow data to a Cloudflare Anycast IP.

Anycast delivers the flow data to a Cloudflare location near the sending data center for analysis. When traffic exceeds attack thresholds, Cloudflare opens an automatic PagerDuty incident for Acme's NOC and begins advertising Acme's prefixes with BGP. Traffic, attack included, reaches Cloudflare within minutes and is blocked at the edge; clean traffic returns through the GRE tunnels with no disruption to end users. Once the attack subsides, the team withdraws its prefixes from Cloudflare with one click and traffic returns to its normal path.

Flow-based monitoring for Magic Transit Embedded Image - 8xGvPP