Find the slow layer. Do not add a cache and hope.
PHP-FPM exhausted, MariaDB waiting on disk, NGINX fine, the theme not our job. We measure the server side of a slow site, then change the part that is actually saturated.
Four layers, looked at in order
A full disk or CPU steal on a small VPS is not a tuning problem. We say if the machine is the limit before editing pool sizes.
Workers, queries, cache. Not the layout.
Pool limits against the RAM you have, a slow request log, opcache actually on. More workers on a 2 GB VPS makes it worse.
Slow query log, a missing index only if the query is yours to change, buffer pool sized to the box. We do not rewrite the application.
Redis for WordPress object cache, NGINX cache for anonymous pages. Cart and admin stay uncached. Varnish only if the site can live with it.
€55 an hour, after a look at what is slow
The first hour is the measurement. Further changes are agreed, not assumed. A bigger VPS is recommended only if the current one is saturated. Moving it is a migration, quoted separately. Theme and plugin rewrites are out of scope.
Site slow, and the host says the server is fine?
Tell us the provider, the RAM and what feels slow. We reply with the first check before any tuning.