Our own ASN, our own address space, our own opinions about silicon.

AMD EPYC 7443P and 9354P for high-density compute nodes, and Ryzen 9 5950X and 7950X for latency-sensitive single-thread workloads. Memory is DDR4 and DDR5 ECC throughout. Storage is enterprise NVMe — Samsung PM9A3 and Micron 7450 class — in RAID-10 on every compute node.

Compute nodes run at a maximum of 3:1 vCPU to physical thread, and typically well below that. Steal time on a healthy node sits under 1%. You can verify this on your own server with vmstat: the st column reports exactly this, and if you see sustained figures above 5% we want to know because it means a node needs rebalancing.

Yes — we announce IPv4 address space along with IPv6 space from our own autonomous system rather than reselling an upstream's. This matters practically rather than symbolically: a provider reselling someone else's addressing can have a customer's prefix withdrawn by its upstream on a phone call, regardless of its own stated policies. Two caveats come with that claim, since we are asking you to check it. The AS number shown on this page is currently AS64511, a number reserved for documentation by RFC 5398, standing in until our registered number is published at launch — do not treat it as a live lookup. The prefix sizes are published at launch alongside the registered ASN. Until the AS number resolves, a prefix size printed here would be a claim you could not check, so we do not print one.

Traffic passes through scrubbing continuously rather than being redirected when an attack is detected, which removes the detection delay during which your service would otherwise be down. Scrubbing capacity sums to 16.5 Tbps across the fleet, but capacity does not pool: the figure that defends you is the one at your own site, which ranges up to 6.5 Tbps and is listed per site in the table above. We do not null-route the customer being attacked — null-routing protects the network by making your server unreachable, which from your side is indistinguishable from the attack succeeding.

Yes, on dedicated servers, at no charge. We establish a BGP session and announce a prefix you hold, or act as upstream if you have your own ASN. You need a valid RIR object and a route object; our network team walks through it in a ticket. Full-table or default-only, your choice.

Open peering at every exchange where we have a presence. We do not require ratio or traffic minimums and we do not charge for peering. Contact the NOC with your ASN and preferred exchange; sessions are usually established within two business days.