A performance-heavy Laravel batch, and the honest kind: as much about what didn’t work as what did, including a Laravel Octane experiment that got measured and reverted.
-
Fast Laravel Case Study: Jason McCreary’s case study is a proper walkthrough of making Laravel News itself faster with plain Cloudflare page caching. The interesting part is that Laravel News looks uncacheable (hourly content, rotating sponsor cards, a newsletter form), and he cached it anyway. Three refactors did it: converting every page from Livewire to Blade components (only the newsletter form actually needed reactivity, so the rest were shedding
wire:state that broke when cached and 419’d), swapping dynamic segments like the sponsor banner to Alpine AJAX fragments that hydrate a cached shell, and purging via model events plus a piggybacked scheduled command for scheduled posts. Result: 15% → 60% cached, edge responses around 20ms. All of it is HTTP caching done systematically, with no paid services. -
Octane vs PHP-FPM: Mattias Geniar’s write-up is the counterpoint I wanted after last batch’s Octane-in-production piece. He moved a fleet of Oh Dear uptime-checker boxes from PHP-FPM to Octane on FrankenPHP, measured real traffic before and after, and moved every one of them back. The reason is clean: Octane’s whole trick is skipping the framework bootstrap per request, and for an I/O-bound workload (each request is a ~400ms outbound check), that bootstrap is a rounding error. Warm workers even pushed idle-box CPU up. The one real win was lower CPU peaks on busy boxes. His rule: Octane earns its keep on CPU-bound, framework-heavy requests served at high rates, not on anything dominated by waiting on the network. Measure your own workload instead of trusting the hello-world benchmark.
-
LaraPlugins Performance Audit: A fellow Romanian, Daniel Petrica, turns his 3AM war story into 12 concrete fixes, all with real numbers. An 8GB VPS pinned at 90% memory, 12 Horizon supervisors with
maxJobs: 0, and a weekly job that fired 50,000 Cloudflare purge jobs in one cycle. The fixes are the good kind of boring: capmaxJobseverywhere so workers recycle memory, a global scope to stop loading a fatpackagist_metadataJSON column on every query, three indexes he’d simply forgotten, and, my favourite, collapsing 50,000 purge jobs into one self-rescheduling job that batches URLs through a RedisSADDset (which dedupes for free). His closing line lands: when monitoring goes noisy you either fix the cause or tune the alerts, and every day you tune the alerts you train yourself to ignore your own dashboard. -
concurrency, parallelism & async in Laravel 13 + Livewire 4: The Developer Awam piece reframes a class of slowness as an execution-strategy problem instead of a database one: the dashboard that calls three APIs in series, the export that blocks for ten seconds. It’s a member-only Medium story so I only read the framing in full, but the distinction it draws (concurrency vs parallelism vs async, and where each fits in a Laravel/Livewire request) is the right mental model, and Laravel’s own concurrency helpers plus queued/deferred work are the tools it points at. Worth it if you keep reaching for a queue when what you actually want is a parallel fan-out.
-
7 Tips to Improve Your ECS Config: Tomáš Votruba’s tips are a back-to-basics cleanup for anyone whose
ecs.phphas sprawled. His line (“if it’s longer than 15 lines, you’re probably doing it wrong”) is the whole ethos: prefer sets over listing individual rules, usewithRootFiles()instead of naming stray files, and lean on the new levels to introduce a coding standard step by step (perfect for agent-driven per-rule PR trains on a legacy codebase). The final config he strives for is three lines of prepared sets plus paths. Feed it to your agent or learn it yourself; either way it’s a good spring clean.