Dedicated Servers for
E-Commerce

Cloud hosting charges per-resource and throttles under load — exactly when your store needs peak performance. An EC2 c5.2xlarge equivalent (∼8 vCPUs, 16 GB) costs €250–300/mo on AWS and shares underlying hardware with other tenants. A bare metal Ryzen 9 9950X (16 real cores, 192 GB DDR5) costs €289/mo in Strasbourg with zero CPU steal and dedicated NVMe I/O. For stores generating €50k+ monthly revenue, the 40–60% infrastructure cost reduction and elimination of throttling during sales events typically pays for the difference within 2–3 months. Shopify Plus (€2,000/mo) makes sense when you need Shopify’s storefront ecosystem; bare metal makes sense when you need full stack control.

For high-traffic WooCommerce: Nginx (not Apache) + PHP 8.2 PHP-FPM with OPcache + MariaDB 10.11 with InnoDB tuning + Redis for object cache and session storage + NVMe for all database files. This stack reduces median PHP response time from 200–400ms (shared hosting) to 40–80ms on bare metal. Add Cloudflare in front for CDN, but keep your origin server on bare metal to avoid the latency overhead of CDN misses on dynamic WooCommerce pages (cart, checkout). If you’re on our Managed Pro plan, we configure this stack for you at no extra charge.

Magento 2’s official minimum is 2 GB RAM, but that’s for a development install with no traffic. In production: Magento PHP processes alone consume 2–4 GB, Varnish cache should be allocated 4–16 GB depending on catalogue size, Elasticsearch or OpenSearch needs 4–8 GB, Redis needs 2–4 GB, and MariaDB/MySQL should have 4–16 GB for InnoDB buffer pool. A mid-size Magento store (50k SKUs, 500 concurrent users) needs 32–64 GB minimum for stable operation. Our Ryzen 9 9950X with 192 GB DDR5 gives all services their recommended allocation with headroom for traffic spikes. For large B2B Magento with complex pricing rules and 500k+ SKUs, the EPYC 9555P with 384 GB is the appropriate baseline.

Strasbourg is our recommendation for most EU e-commerce stores: GDPR data residency in France, 170 km from Frankfurt with equivalent EU routing, and the lowest entry price for our Xeon and Ryzen 9 configurations. Frankfurt is the better choice if your primary customer base is Germany, Austria, or Switzerland (D-A-CH), or if you need DE-CIX peering for sub-8ms latency to German users. Both locations reach London in under 15ms and Amsterdam in under 12ms. For stores serving both EU and North American customers, add a second server in Miami or New York — the multi-region combination starts at €408/mo.

Yes. Bare metal handles traffic spikes better than shared cloud because there’s no CPU throttle ceiling. The server’s full resources are always available — a 10× traffic spike during BFCM hits the same hardware that handles your baseline load. Best practice for BFCM preparation: verify your Varnish hit rate is above 90% for category/product pages, pre-warm the Elasticsearch index, increase PHP-FPM max children to match expected concurrency, enable Nginx microcaching for non-authenticated pages, and run a load test 2 weeks before the event. If you’re on Managed Pro, we’ll run a pre-BFCM health check and tuning session at no extra charge.

Not on bare metal. The reason agencies separate Elasticsearch onto its own VM is resource contention on shared hosting — Elasticsearch’s 4–8 GB JVM heap competes with PHP-FPM and MySQL on a 16 GB VPS. On a Ryzen 9 9950X with 192 GB DDR5, all services have their full RAM allocation without competition. Running Elasticsearch on the same bare metal host as Magento or WooCommerce is the standard architecture used by high-volume bare metal deployments. We allocate heap via JVM config during initial setup if you’re on a Managed plan.