On-Premise Deployment

Our filtering,
inside your
datacenter.

For networks that cannot route traffic through someone else's cloud — sovereignty rules, data residency, regulated workloads, or contracts that simply forbid it.

Topology

Three ways to put it in the path.

Which one you want depends on how much steady-state latency your traffic tolerates and how quickly mitigation has to engage. Not every engine behaves the same way in each — where it differs, we say so.

InternetOriginSCRUB

Always-on

Inline

Appliances sit in the forwarding path. Every packet is inspected on the way through.

Trade-off

Sits in the path, so it deploys in redundant pairs with hardware bypass.

InternetOriginSCRUBdiverted on detect

On-demand

BGP diversion

Traffic flows normally until detection fires, then the targeted prefix is diverted through scrubbing.

Available
  • Pyro

    Drops established connections when protection engages.

  • Floe

    Drops established connections when protection engages.

  • Cryo

    Drops established connections when protection engages.

Trade-off

Mitigation starts once diversion converges, not at the first malicious packet.

InternetOriginSCRUBoverflow to edge

Hybrid

Inline + edge overflow

On-premise appliances absorb what they can; overflow is announced to the Papyrus edge.

Trade-off

Requires an agreement to announce your prefixes during an overflow event.

What you get

Built for networks that answer to a regulator.

The same engines, on your metal

Cryo, Pyro and Floe ship as appliance images rather than a separate on-premise product line. A signature that lands at the edge lands here in the same release.

Runs air-gapped

No phone-home requirement. Signature and ruleset updates can be delivered as signed offline bundles for networks that permit no outbound management traffic.

Traffic never leaves your estate

Packets are inspected and dropped inside your own facility, which is what makes this viable under data-residency and sovereignty obligations.

Line-rate filtering in kernel

Filtering runs in eBPF/XDP before the network stack, so drop decisions cost a fraction of what a userspace proxy costs at the same packet rate.

Logs stay yours

Flow records and mitigation events export straight into your own SIEM. Nothing is retained on our side unless you ask us to co-manage.

Your NOC or ours

Run it entirely in-house, hand operations to our engineers, or split it — escalation paths are defined per deployment rather than sold as a fixed tier.

Start with a design review.

On-premise deployments are scoped, not ordered. Tell us what you run and where, and an engineer will come back with a topology and an appliance count.

Talk to an engineer
  1. Design review

    We walk your topology, transit mix and traffic profile with a network engineer.

  2. Sizing

    Appliance count and placement are derived from your real packet rates, not a plan tier.

  3. Staged rollout

    Bypass-first install, then progressive enablement per prefix with rollback at every step.