Performance, uptime & infrastructure
How this is scored
What runs the sites and how reliably: CPU, RAM and storage per tariff stated, SSD or NVMe storage, current PHP versions and how long old ones are kept, HTTP/2 or HTTP/3, server-side caching (OPcache, Redis, a page cache), a CDN, databases and their limits, resources that are guaranteed rather than shared without a stated limit, an uptime SLA with a service credit, a public status page with incident history, and redundancy of the platform. Where the data centre is located is scored under sovereignty.
0 — Not a hosting product; websites or web apps cannot be run on it.
3 — Storage and a number of sites per tariff are stated and the product is described as fast; resources, PHP versions, caching and uptime are basic or not described.
5 — Resources per tariff stated (CPU or a performance class, RAM, SSD or NVMe storage), several current PHP versions selectable, HTTP/2, server-side caching or a cache plugin documented, free SSL, a stated uptime target, and a public status page.
8 — Dedicated or guaranteed resources stated per tariff, HTTP/3 and a CDN included, Redis or an object cache, PHP versions kept current with an end-of-life policy published, an uptime SLA with a service credit, a status page with incident history, and a redundant platform (replicated storage, failover) described.
10 — Infrastructure documented to the level an operator could audit: hardware generations and storage architecture named, performance measured and published with method, an SLA with credits and exclusions at 99.95 % or above, every incident on the status page with a post-mortem, and scaling without downtime documented.
The Shop Operator
Guaranteed RAM up to 8 GB, NVMe storage up to 300 GB per package, a PHP version per domain with PHP-FPM and free SSL give me a clear resource picture, and the Hamburg data centre is documented down to redundant disks, power supplies and switches, 15 minutes of UPS buffering, emergency power for at least 24 hours and 24/7 on-site monitoring. For a Black Friday weekend I still miss the operational proof: the captured pages give different figures for the guaranteed uptime — at least 99.5 % on the agency page and at least 99 % in the annual average in the terms — and we found no public information on a service credit, a status page with incident history, Redis, or HTTP/2 and HTTP/3. 1 4 6 10 11