So I got this "fast" vps from OVHcloud for €18/year and the wordpress bench says 2000 req/s but actual woocommerce checkout? 8 seconds. Definitely not the same thing. Anyone actually tested real cart flow or is everyone just looking at hero numbers
Real-world WordPress benchmark vs synthetic hero numbers
🚫 mas tipo checkout é cache é redis? Object cache?
Ah~ actually I had same problem with Hetzner. Basically woocommerce is heavy, no? (´・ω・`) I added redis object cache and nginx fastcgi cache~ then checkout went from 6s to 1.2s. But honestly static html with webhook for inventory is more stable~
For what it's worth, the synthetic benchmark measures PHP throughput under ideal conditions. WooCommerce checkout involves database writes, session management, and often third-party API calls for payment processing. Hot take: your "2000 req/s" is meaningless if 80% of requests then hit a 30-second gateway timeout on checkout. Check if your provider's MariaDB is on NVMe or spinning rust—that's usually the chokepoint, not the vCPU.
So I run my shop on a gemütlich three-node k8s cluster at Vultr (https://www.vultr.com) and the CheckoutFlow is actually three microservices with a redis-cache-layer. Na ja, the Database connection still bottlenecks sometimes. My hot take: the WordPress benchmark numbers are Marketing numbers, not Reality numbers. You need ObjectCaching and probably a CDN for the static Assets.
Ok so heres the thing larry I did the exact same dance with InterServer last year,,,,, Synthetic numbers looked great cart was a dumpster fire,,,,, Ended up just running static html with a webhook to update stock when someone buys,,,,, The scrap gods provided an old dedi for $15/mo and I just generate the whole site every 5 minutes,,,,, No php no mysql no tears,,,,, ~~dave
8 seconds for checkout? Even on that $18 plan that seems rough.
What php version and mysql are you running on OVHcloud
Same dumpster fire with InterServer. Synthetic numbers mean nothing for carts.
OVHcloud doesn't exactly publish that for the €18 tier but ioping says 2.3ms so probably not NVMe. Good call.