Purpose-built infrastructure. Not a rented dashboard.
Novixa runs on hardware we chose, in facilities we vetted, with long-term commitments to the operators who run them best. No virtual resale. No mystery uplinks.
A delivery network is only as good as the metal underneath it. The industry's bad habit is to rent a VM, tag it with a brand, and call it a PoP. We don't.
Novixa operates full-cabinet colocation at each of our core sites — with physical access, dedicated cross-connects, and hardware we refresh on a three-year cycle. In the United States, we partner with SpartaHost for full-cabinet hosting across their Tier III facilities. In Asia, we operate in carrier-neutral exchanges in Hong Kong and Tokyo that terminate CN2 circuits directly.
This is how we guarantee the network. When we tell you a zone is fronted from HKG, we mean HKG — the cabinet, the switch, the line card. Not a re-advertised route from three hops away.
Every Novixa core site is a facility we hold cabinets in — not a reselling POP.
Our US footprint is operated in partnership with SpartaHost across three coast-to-coast Tier III facilities. Full-cabinet hosting, private cross-connects, and dedicated upstreams on Cogent, Telia, and NTT.
- Tier III+ facility certification
- A/B power feeds, N+1 cooling
- Dual 100G carrier uplinks per cabinet
- 24/7 staffed NOC access
Our APAC sites are built in carrier-neutral exchanges that terminate CN2 GIA, CN2 GT, CMI, and CUG circuits directly. This is where the China-optimized routing lives — and where most of our Asian customers' traffic is served.
- CN2 GIA + CN2 GT circuits on-net
- CMI and CUG private interconnects
- Dual-path diversity through separate cable landings
- Sub-1 ms intra-region latency
European sites peer directly at the continent's three largest internet exchanges. Over 1,200 networks are a single AS-hop away, keeping European origins fast for European audiences.
- Direct peering with 1,200+ networks
- 100G IX uplinks per site
- DDoS scrubbing co-located with edge
Standardized across every site, refreshed on a 36-month cycle.
Dual-socket AMD EPYC 9004 series with NVMe-attached local storage. Sized for the edge — high clock speed and memory bandwidth, not core count for its own sake.
Arista / Juniper top-of-rack switching with 400G uplinks. Every cabinet is dual-homed; every link is BFD-monitored with sub-second failover.
Local NVMe for hot cache. Regional object storage on erasure-coded clusters with per-object checksums. No single-point-of-failure caches.
Full packet capture on capture nodes, flow data on every transit link, and per-request logs shipped to our lake within 60 seconds. You see what we see.
A network you can actually audit.
We'll share facility letters, cross-connect diagrams, and real-time BGP advertisements. Ask.