Skip to content
Core Web Vitals specialists

Server & hosting

Before the browser can display anything, the server has to respond. This category covers everything before the first line of HTML: server choice and resources, time to first byte, database queries and their indexing, PHP configuration and OPcache, HTTP versions and connection setup, compression with Brotli and Gzip, and caching at several levels — from object caches to full page delivery. We look at delivery through content delivery networks, at sensible expiry times for static files and at when a migration is worthwhile. Described so you can talk to your host or IT team about causes in concrete terms. We additionally cover which details you need from a provider to compare offers objectively instead of deciding by package name.

AI crawler load: where the requests landHTML requests and crawl purposes, June 2026HTML requestsJune 2026Bots57.5%Humans42.5%Cloudflare RadarWhere bot requests end upBot requestCacheable: home, categoryLong tail: filters, searchOrigin: PHP workers, DBHitMissAI crawls by purpose52%TrainingTraining 52%Mixed use 36%+Cloudflare RadarCapacity levers: throttle instead of blockFacets and parametersremove from the crawl pathRate limiting at the edgeprioritize, do not banCache rules for botsmake long tail cacheableProtect the worker poollimits against exhaustionCrawl-to-referral per bot shows which crawler creates load and which one sends visitors (Cloudflare Radar)
Server & hosting Jul 22, 2026

AI Crawler Load: When Bots Slow Down Your Server

Bots generate over half of all HTML requests. Why AI crawlers hit uncacheable long-tail URLs, how to measure the load and throttle it instead of blocking.

14 min read
Text Compression 2026: Fewest Bytes, Fastest ResponseMobile HTML compression (Web Almanac)Gzip52%Brotli37%none11%Brotli on 37% of pages, 11% with no compression at allCompression throughput (Zstandard project)zstd -1510 MB/sGzip -1105 MB/sabout 5x faster compression than Gzipdecompression stays fast at every levelStatic: Brotli 11 smallest files | Dynamic: zstd about 4x faster | Fallback: GzipStatic build assetsBrotli level 11smallest file, compressed onceDynamic HTML / JSONzstd level 12-19fast, low CPU loadOld clients / HTTPGzipuniversal fallbackContent-Encoding negotiation and fallback chainClient sends Accept-Encoding: zstd, br, gzip -> server picks best shared method, otherwise GzipChrome 123+Edge 123+Firefox 126+Safari 26+(Can I use)
Server & hosting Jul 15, 2026

Brotli or Zstandard: Faster Web Compression 2026

Brotli, Zstandard or Gzip for web compression in 2026: which method delivers the smallest transfer size and fastest response - static and dynamic alike.

12 min read
Query Waterfall: Unindexed N+1 vs. Indexed and CachedDatabase time of a category page with 20 productsbefore: N+1 without indexSELECT list 18 ms+ 20x SELECT per productfull scan, no indextotal DB time 420 msafter: eager load + indexSELECT list 9 ms1x JOIN products (index) 6 msquery cache HIT 0.3 mstotal DB time 15 msQuery time before and afterbefore420 msafter15 ms0.8 sgood TTFB value (web.dev)1%revenue per 100 ms (Amazon)8.4%more conversions per 0.1 s32%more bounce 1 to 3 sA good TTFB is 0.8 seconds or less (web.dev) -- every additional 100 ms of latency cost Amazon about 1 percent in sales (Amazon).A mobile load time 0.1 seconds faster increased retail conversions by 8.4 percent (Google/Deloitte 2020).Illustrative values for the N+1 pattern; real timings vary by data volume and hardware.
Server & hosting Jun 15, 2026

Database Query Optimization for Shop Performance

How indexes, resolving N+1 queries, query caching and EXPLAIN speed up slow database queries and lower the Time to First Byte of your online shop fast.

14 min read
Edge Delivery: Origin Alone vs. CDN with Edge PoPsTime to First Byte by distance to the userOrigin serverFrankfurtEdge PoPNew YorkHIT 12msuserEdge PoPSingaporeHIT 15msuserEdge PoPSao PauloMISS 14msuseron MISS onlyorigin fetchTTFB for a user in Sao Paulowithout CDN320 mswith edge HIT15 ms54%requests via CDN1% revenuelost per 100 ms90.4%responses cacheable124 mi/mslight in fiber54 percent of all requests are served from a CDN (Web Almanac 2024) --every additional 100 ms of latency cost Amazon roughly 1 percent in sales (Amazon).
Server & hosting Jun 10, 2026

CDN and Edge Caching for Shop Performance 2026

How a CDN, edge caching and well-set cache-control headers cut Time to First Byte worldwide and keep your shop fast even far from the origin server.

14 min read
Hosting Model Sets the Server ResponseShared Hosting800-2000 mstypical TTFBshared resourcesfragile under loadVPS (self-managed)300-700 mstypical TTFBdedicated capacityscaled manuallyManaged Hosting200-500 mstypical TTFBoptimized stackcache built inCloud / autoscaling100-400 mstypical TTFBscales elasticallyedge-distributedsetsTTFB caps the LCP budget2.5 s (good)0 sRender delay 10%Resource load time 40%Load delay 10%TTFB - server response 40%set by hostingGood TTFB at the 75th percentile: under 800 msOnly 42% of mobile sites reach it (HTTP Archive) - frontend cannot lift the ceiling
Server & hosting May 11, 2026

Hosting Choice: How the Server Shapes Load Time

The server sets the ceiling for load time. How shared hosting, VPS, managed and cloud differ in TTFB, load spikes, scaling and latency to your audience.

13 min read
Back to the blog overview