Neurowall edition · ISPs, datacenters & hosting providers

NeuroWall Edge.
Built to sit
at the edge of
someone else's network.

Most firewalls protect one network: yours. NeuroWall Edge protects many networks at once — each belonging to someone else — from a single box you operate. It's the datacenter gateway deployment plus a self-service tenant API, packaged for ISPs, datacenters, and hosting providers.

FORWARD
Primary enforcement path
No DNAT
Routed public prefixes
3/4/5-tuple
Tenant-owned rules
404
On cross-tenant access, not 403
The deployment

A gateway for public prefixes, not just a box with a public IP.

Most firewall products assume they're protecting the host they're installed on, or a private LAN behind NAT. Neither assumption holds at datacenter scale.

01 · Topology
Routed public prefixes, not NAT

A datacenter or ISP gets one or more public prefixes routed to a gateway by its upstream. The gateway forwards traffic straight through to servers carrying real, routable public IPs — no DNAT, no port-forwarding gymnastics.

02 · Enforcement
Policy lives in the FORWARD path

Datacenter gateway mode enforces policy primarily via nftables in the FORWARD chain — WAN to servers, servers to WAN — with eBPF/XDP handling the fast-path blocklist and DDoS layer underneath it.

03 · Fit
A third deployment mode, not a retrofit

Datacenter gateway mode sits alongside NeuroWall's existing host-firewall and NAT gateway modes — built for the "routed public subnets behind a gateway" shape ISPs and datacenters actually run.

The customer-facing half

A tenant API, not a ticket queue.

A gateway only the provider can touch solves half the problem. The other half: what happens when the customer needs to react to an attack right now?

Tenant self-service

Name, subnet, API key

The provider provisions each customer as a tenant: a name, assigned IP space, and a tenant API key. Customers can then manage permitted rules without routing every change through the NOC.

Scoped to owned IPs

3/4/5-tuple block and allow rules are checked against the tenant's assigned subnet(s).

ISP Customer Portal

A dedicated workspace for customers

The separate portal connects customer accounts to registered NeuroWall instances. Customers can review their assigned network and firewall rules; provider administrators manage customers, bindings, and gateway registrations.

The portal keeps its own accounts and database. Its server communicates with NeuroWall's tenant API, keeping tenant credentials out of browser code. Customer access and provider administration remain separate.

See the portal

A clearer view of customer-side operations.

See how the ISP Customer Portal brings gateway registrations and customer operations into one workspace.

If the embedded player is unavailable, watch the video on YouTube.

Isolation guarantees

What that isolation actually guarantees, not just promises.

  • 01
    A tenant can only ever touch their own IPs.
    Every rule a tenant tries to create is checked against their assigned subnet(s) before it's allowed to exist. An attempt to write a rule against another tenant's IP space is rejected outright.
  • 02
    A tenant can't even see that other tenants exist.
    Listing rules or subnets returns only that tenant's own resources — there's no window into anyone else's slice of the network, deliberately, not just by omission.
  • 03
    Deleting someone else's rule looks like it was never there.
    A tenant trying to delete a rule that belongs to another tenant gets a 404, not a 403 — the API won't even confirm the rule exists, let alone whose it is.
  • 04
    The provider keeps a completely separate front door.
    Tenant authentication and provider/admin authentication share no middleware, no context, no failure path.
Why one edition, not two features

Neither half is much use without the other.

A fast gateway only the provider can operate still routes every "block this IP" request through the same 2am ticket queue. A slick self-service portal with nothing fast underneath it just moves the bottleneck from "waiting on support" to "waiting on the firewall to catch up."

Kernel-speed enforcement

For the traffic pattern datacenters and ISPs actually have — routed public prefixes, no DNAT, policy in the FORWARD path.

Immediately-effective self-service

For the customers living on top of it, with isolation strict enough that a provider can hand over control without inheriting new risk in return.

FAQ

Frequently asked questions.

  • NeuroWall Edge is a Linux-based, multi-tenant firewall and DDoS protection platform designed for ISPs, datacenters and hosting providers. Powered by eBPF/XDP, it enables high-performance traffic filtering, destination-based DDoS protection and customer-specific firewall policies. Tenants can manage security rules for their assigned IP addresses and subnets without affecting other customers.
  • NeuroWall Edge is deployed on Linux gateway infrastructure within an ISP or datacentre network, typically along the traffic path between upstream connectivity and customer networks. It can operate as a transparent gateway or routed firewall, applying eBPF/XDP filtering before traffic reaches protected servers. Multi-tenant policies isolate customer configurations while allowing centralized operator management.
  • NeuroWall Edge's throughput depends on server hardware, NIC capabilities, CPU performance, packet size and enabled security policies. Its eBPF/XDP architecture supports high-speed packet processing on compatible Linux network interfaces. Actual supported link speeds and packets-per-second capacity should be established through hardware-specific benchmarks rather than assuming a fixed throughput limit.
Who this is for

Infrastructure for other people, not just for yourself.

A datacenter with routed public prefixes. A hosting provider. An ISP handing customers their own IP space. If "please block this IP" is a recurring line item in your support queue, this is the edition built for exactly that shape of problem.