PhantomInfrastructure
What "reliable" is made of
Uptime isn't a promise, it's a bill of materials: the machines we run, the routes we peer, the filters we build, and the people we wake up. Here's ours, itemized.
The hardware
Current-generation, tested, and retired on purpose
Every server in the Phantom fleet is spec'd, configured, and managed by our own team, on standardized builds across our footprint. When something needs fixing, the people fixing it already know the machine — no waiting on a third party's ticket queue, no guessing at the build.
We run current-generation silicon and cycle hardware out proactively at the end of its performance life, not its legal life. Drives are monitored and swapped on wear thresholds before they fail; memory runs ECC wherever the workload can use it; and every machine goes through burn-in testing before it carries a customer byte.
Standard fleet today: AMD Ryzen 9 7950X for high-clock game workloads, AMD EPYC 9354 and 9554 for density and virtualization, and Intel Xeon Scalable for specific enterprise and compatibility requirements. All storage tiers are flash-first.
The network
Latency is a route, and we choose ours
Packets don't care about marketing; they care about hops. We select and manage upstream network partners for extensive direct peering — 340+ peering relationships in total — with the ISPs and exchanges where our customers' users actually live. The fewer networks between your server and your player, the fewer things can go wrong in between.
We maintain a presence in twelve locations worldwide, each backed by redundant upstream transit from independent carriers, diverse fiber paths, and automatic rerouting that converges in under thirty seconds. Congestion is engineered around before it happens — capacity is upgraded at 40% sustained utilization, not 90%.
The result is the number on our homepage: a median under 20 ms to major population centers on four continents, measured continuously, not quoted from a launch-day test.
The shield
DDoS mitigation you learn about afterward
Denial-of-service attacks are a weather condition in this industry — especially around game communities, where a lost match can motivate a rented botnet. We treat mitigation as core infrastructure, not an upsell: every service sits behind multi-terabit filtering from the moment it's provisioned.
What makes ours different is protocol awareness. Game traffic has patterns — small packets, high frequency, latency sensitivity — that generic scrubbing mangles. We build and tune per-game filtering profiles, so legitimate players pass untouched while amplification floods, spoofed SYNs, and application-layer abuse are dropped at the edge, before they ever reach your server.
Most attacks are absorbed with zero customer-visible impact. You'll find them itemized in your monthly report, usually as a surprise.
The commitment
99.99%, with consequences for us
Our uptime SLA covers network and host availability at 99.99% monthly. If we miss it, service credits apply automatically — calculated by us, issued by us, without requiring you to notice the outage, open a ticket, and argue. An SLA you have to enforce yourself isn't an SLA; it's a scavenger hunt.
Behind the number: our facilities run on N+1 power and cooling, our hypervisor clusters are redundant with live migration for maintenance, our monitoring checks every service from multiple continents every ten seconds, and our incident process publishes a first update within fifteen minutes — even when the update is "we're still investigating."
Maintenance windows are announced at least seven days ahead, scheduled for regional low-traffic hours, and designed to be non-disruptive. Most complete without a single dropped connection.
By the numbers
The fleet, measured
Put it to work
This much infrastructure, starting at $15.50 a month
Everything on this page sits underneath the smallest VPS and the largest enterprise cluster alike. Pick a shape and deploy.