We received an ASN and IPv6 space for GrapheneOS from ARIN: AS40806 and 2602:f4d9::/40.
We've deployed 2 anycast IPv6 networks for our authoritative DNS servers to replace our existing setup: 2602:f4d9::/48 for ns1 and 2602:f4d9:1::/48 for ns2. BGP/RPKI setup is propagating.
We applied for an IPv4 /24 for ns2 via NRPM 4.10 and can apply for one for ns1 after we obtain that one.
Our ns1 network has New Jersey, Miami, Los Angeles, Seattle, Frankfurt and Singapore. Our ns2 network currently has New York, Las Vegas and Bern. We'll be expanding both.
This provides an overview of worldwide latency for our ns1 cluster via the Rage4 anycast service we currently use for IPv4+IPv6 with ns1:
ping6.ping.pe/2a05:b0c4:1::8
Here's ns1 via our own IPv6 /48:
ping6.ping.pe/2602:f4d9::1
Here's ns2 via our own IPv6 /48:
ping6.ping.pe/2602:f4d9:1::1
In the future, we plan to use these 2 anycast networks to provide recursive DNS resolvers as an option for our users. For now, it's only for the authoritative DNS used to provide other GrapheneOS services which is what DNS resolver servers query after the root and TLD servers.
ARIN gave us an IPv4 /24 based on our NRPM 4.10 request in under 24 hours. It's being announced from our ns2 network:
github.com/GrapheneOS/ns1.grap…
It will take a long time to propagate since the RPKI IRR/ROA data gets fetched via timed jobs rather than pushed hop-by-hop like BGP.
It cost us US$50 to register with ARIN as an organization and US$262.50/year paid in advance to become an 3X-Small network. It'll be US$525/year when we get a 2nd IPv4 since we'll get pushed into 2X-Small. 2X-Small covers IPv4 /22, i.e. 4x /24, which we can get via the waitlist.
We've deployed our IPv4 /24 and IPv6 /48 for ns2 in production to replace the IPv4-only anycast tunnel system it relied on before. It has somewhat better latency and significantly better reliability now. We're waiting a bit longer for production deployment of our ns1 IPv6 /48.
We need to choose a host in Singapore with IPv4+IPv6 BGP support to extend ns2 with a location in Asia. Once that's added, it will be good enough for our current needs. The subset of our dedicated/colocated update servers with BGP could be used as extra ns2 locations eventually.