Journal
Performance
Notes in Performance — written like an editor, not a keyword farm.
Performance · August 16, 2026
PHP-FPM settings that actually move the needle
Default PHP-FPM pools are conservative. On a busy WordPress or Laravel site they become the bottleneck long before CPU is fully used. A few settings, changed carefully, often produce a visible improvement. The ones that matter most pm and pm.max_children: Too low and requests que…
Performance · August 8, 2026
HTTP/3 and QUIC: the real gains on shared and VPS hosting
HTTP/3 (running over QUIC) is no longer experimental. On stacks that support it—LiteSpeed is a common example on shared hosting—the gains are measurable, especially for mobile clients and lossy networks. Where the improvement shows up Faster connection establishment (one round tr…
Read
Performance · August 7, 2026
When origin location still beats the CDN
CDNs are excellent for static assets and cacheable pages. They are less magical for personalized or highly dynamic responses that must come from the origin on every request. In those cases the physical location of the origin still dominates latency. When the origin matters most L…
Read
Performance · August 6, 2026
NVMe is table stakes — the latency numbers that actually matter in 2026
Every serious VPS and most shared platforms now ship NVMe. The storage medium is no longer the differentiator it was five years ago. What still separates fast sites from slow ones is the distance to the user, the application architecture, and the consistency of the host’s network…
Read
Performance · August 5, 2026
Measuring real-world TTFB across six cities
LogicWeb operates VPS in New York, London, Frankfurt, Singapore, Tokyo, and Tel Aviv. The useful question is not which city is “fastest” in the abstract. It is which city is fastest for a given audience. Why city choice still dominates Light travels fast but not infinitely fast. …
Read