Seven applications benchmarked on one 2 vCPU VPS. Six land between 11 and 43 requests per second per core regardless of language or framework, and caching is the only lever with an order of magnitude in it. With the isolation model, the economics with your own hours priced in, and the application-layer incident that mattered more than the perimeter.
Hardening steps ordered by measured effect, from a live server that logged 1,422 failed SSH attempts in one week. Includes the two silent fail2ban failures we found on our own machine, and the sysctl audit showing nine of ten hardening settings are already Ubuntu defaults.
We benchmarked the open-source alternatives instead of guessing. Shlink and Invoice Ninja on an entry-tier VPS: 22.8 and 19.1 requests per second, against 190.8 for a minimal Laravel route on identical hardware. Application complexity, not the framework, sets your server sizing.
Cloudways Medium is $54 a month for 4 GB and 2 vCPU; a Hostinger KVM2-class VPS is around $9 for 8 GB. The decision is labour, not hardware: the price gap buys 52 minutes of your month. With measured VPS benchmarks showing a 40x spread from caching alone, and a protocol for benchmarking any host yourself.
Both deployed on a live VPS under identical 1 vCPU limits. Cal.com idled at 1.05 GB of RAM with an 8.05 GB image; Easy!Appointments at 27 MB and 877 MB. Includes the reverse-proxy rate-limiter trap that returns 429 to every visitor, and its verified one-line fix.
We ran the usual eval and base64_decode checklist against 7,441 PHP files and 1,674,891 lines of production code. It produced 588 hits, all benign, and missed a four-line backdoor entirely. Here is the layered version that reduced the corpus to 11 files worth reading.
Measured on a live n8n 2.37.10 deployment limited to 1 vCPU: 11.4 executions per second, a 402 MB idle footprint, 16.9 second cold start, and 3,563 bytes of database growth per execution. Complete Compose file, nginx reverse proxy, queue mode, and five documented failure modes.
Measured on a live 2 vCPU KVM VPS running Laravel 13.30.1: 16.4 req/s without OPcache versus 190.8 req/s fully tuned. Complete nginx, PHP-FPM and systemd configuration, an atomic zero-downtime deploy script, and five documented failure modes.
Measured on a live 2 vCPU KVM VPS: 19.9 req/s uncached versus 749.6 req/s with an nginx FastCGI microcache. Complete nginx and PHP-FPM configuration, a marketplace script security audit protocol, and four documented failure modes.
↑↓ navigate ↵ open esc close