Unbox™.
Infrastructure is not a commodity—it is a foundation. Unbox™ is our documentation project to expose the reality of our stack. From the specific NVMe drive models in our server clusters to the custom routing logic in our edge network, we believe transparency is the first step toward reliability.
The physical layer.
We don't just lease white-labeled racks. We partner with Tier-4 data centers to deploy our own hardware clusters. Every Starlight™ VM is backed by redundant arrays that prioritize IOPS over capacity, ensuring that your database transactions never wait on disk latency.
Dual-socket AMD EPYC™ 7003-series processors. High core density allows us to keep VM resource allocation strict and predictable.
All instances run on Samsung PM9A3 NVMe Gen4 drives in RAID-10 configuration. We sacrifice storage density for read/write longevity and error correction.
Connectivity topology.
Our network is built on a BGP-anycast backbone. When a user requests your site, our edge nodes detect their geolocation and route them to the nearest available data cluster, minimizing the round-trip time (RTT) for TLS handshakes.
The Unbox Philosophy
In an industry dominated by "abstracted magic," we prefer the cold, hard reality of circuit boards, kernel versioning, and latency metrics. Unbox™ is our commitment to providing you with the exact technical specifications of every service we sell. You are not buying a "cloud solution"; you are buying compute, memory, and bandwidth on hardware we manage with surgical precision.