The cheapest cache is the work you do not repeat
Before you install Redis, ask how many times a day the data actually changes.
On a personal site, most content changes a few times a month. It is read thousands of times. Anything computed on every request is work paid for a thousand times and used once.
Three simple moves
Parse on write, not on read. Markdown becomes HTML when you press Save. The front end serves finished HTML.
Settings in a single file. All settings become a var_export written to disk. One include, zero queries, and OPcache already holds it compiled.
Sitemap and feed on disk. Regenerated hourly, not on every crawler visit.
What I did not do
I did not cache the article pages. They load in a few milliseconds, and the first time you add a caching layer you also add a whole class of bugs: stale content served with confidence.
A wrong cache is harder to debug than a slow page.
