Peak-hour profile · Madrid
Madrid hosting, measured at the hour it matters
Your traffic peaks at ten in the evening, your tests all ran at eleven in the morning, and the two numbers have never been in the same room.
The short answer
Measure at the hour your audience actually arrives. Spanish consumer traffic runs late, so a lab test on a quiet Tuesday morning tells you nothing about the evening peak — and the fix for a peak is cache ratio rather than raw capacity, because a cache hit is answered without the application booting at all.
The platform side is fixed and generous: NVMe storage behind a LiteSpeed cache built into the server on every plan from $2.42/mo, OPcache on current PHP 8.x, and a published 10–35 ms round trip from Iberia to our London origin. Free SSL, free migration and engineers on shift whatever the hour.
By the HostingFast team · Reviewed 12 August 2026
Free
Migration, engineer-run
24/7
Engineers on shift
Daily
Backups, restore yourself
$0
Setup, every tier
Banks and utilities are headquartered along the Castellana, television and film studios feed an audience of half a billion, and a service economy opens late and stays open later. The traffic shape here is genuinely different from northern Europe, and it changes which numbers are worth reading.
This page covers the peak, the payload that media sites carry into it, what splitting a Spanish-language site actually costs, and where 10–35 ms shows up at all.
Measure at ten in the evening
A synthetic test is a snapshot and you choose when it happens, so choose the hour that matters. Run the same three curl commands against your busiest URL at the peak and again at a quiet hour, and compare. If the peak figure is far worse, the site is running out of capacity on uncached requests rather than being slow by design.
Then read the field data, which does not care what time you tested: PageSpeed Insights reports real Chrome visits at the 75th percentile, aggregated across the whole day, and the Core Web Vitals thresholds are assessed there — LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1. Lab tells you the mechanism; field tells you the audience.
A peak is a cache-ratio problem before it is a capacity one
If ninety per cent of evening traffic is people reading, ninety per cent of it can be answered pre-built without the application booting. That is why cache ratio, not core count, is the first thing to fix: request your ten busiest URLs twice each and record which return a hit on the second attempt. The ones that never hit are where the peak actually hurts.
The usual blockers are a session cookie set sitewide, a personalisation block in the header, or a comment widget loading on every article. Fix them on staging, re-run the same ten URLs, and only then consider the tier above. Nitro's 2 cores and 4 GB are the right answer for a genuinely uncacheable workload, and the wrong answer for a caching misconfiguration.
Media payloads, weighed rather than admired
Television and publishing sites carry the heaviest assets, and the largest contentful paint is almost always the lead image or a video poster frame. WebP conversion is built in, so the format is handled; what remains is the pixel count, the number of font weights, and whether an ad or analytics stack executes before the headline paints.
One rule saves more evenings than any other: never mark the lead image lazy. Deferring the element the browser is timing is the most common self-inflicted regression on an editorial layout, and it looks like an optimisation on every checklist that recommends it. Set width and height attributes instead and keep CLS under 0.1.
One Spanish site, or five, and what the difference costs
Start with one site in es-ES with hreflang declared, and spin out an es-MX or es-AR variant only when the pricing, the offer or the vocabulary genuinely part company. Every split doubles the certificates, the backup sets, the redirect chains and the cache entries you have to keep correct, in exchange for a difference your customers cannot see yet.
The technical side is straightforward either way: .es converts at home and its DNS points here like any other name, registration sits on the same account as the hosting, and SSL issues as soon as each name resolves. What matters is keeping one canonical hostname per site and a single 301 to it — verify with curl -sIL rather than trusting the configuration.

What holds up at the peak
Server-level LiteSpeed caching on NVMe storage with OPcache on current PHP 8.x, from $2.42/mo upward — so the evening crowd is answered from pre-built pages rather than from the database.
Plus the instruments: cache status in the headers, staging copies, per-site PHP versions, and daily backups you restore from the panel yourself.
- Cache hit or miss readable from every response
- WebP conversion built in at no extra cost
- .es registered and renewed on the same account
- Staging copies to test a change at peak-hour load
Why HostingFast
Standard on every plan
Pre-built pages at the peak
A cache hit is answered without the application booting, which is what keeps a ten o'clock crowd from becoming a capacity problem.
Format conversion included
Built-in WebP means a media library serves smaller files to modern browsers without anybody adding a build step.
Cheap uncached requests
NVMe storage with OPcache on current PHP 8.x keeps the pages that cannot be cached from dominating the evening.
Headroom when it is genuinely needed
Nitro carries 2 cores, 4 GB, 200 GB of NVMe and backups every 6 hours — the right answer once the cache ratio is already good.
Zero setup, flat renewal
No joining fee at any tier, and a renewal at the rate you ordered at, so the annual figure does not move under a media budget.
Migration without downtime
Engineers move the site free while it keeps serving, and delegation changes at a time you pick rather than one we impose.
Quick Start
From order to online
- 1
Test at the peak hour
The same three curl runs at 22:00 and at 11:00 against your busiest URL. If those two numbers differ a lot, capacity is being spent on uncached requests.
- 2
Raise the cache ratio
Ten busiest URLs, two requests each, and fix whatever never returns a hit. That is cheaper and more effective than any tier upgrade.
- 3
Weigh the lead assets
Export dimensions, font weights, and whether the lead image is lazy-loaded. Then wait for the field data to confirm a visitor felt it.
Built In
Loaded onto every plan
- LiteSpeed caching built into the server rather than patched in by plugin
- Built-in WebP image optimisation at no extra cost
- NVMe SSD storage on every tier, entry plan included
- Current PHP 8.x with OPcache enabled from the start
- Staging environments for testing changes before they ship
- Per-site PHP version switching from the control panel
- Daily backups, with self-service restores you run from the panel
- Free SSL on every plan, renewed automatically before it can expire
- Zero setup charges — there is no joining fee, ever
- 99.9% uptime as the target, monitored around the clock
Frequently Asked
What people ask us most often
Why does the site feel slower in the evening than in the morning?
Because the morning test was answered from cache and the evening one was competing for capacity on uncached requests. Run the same curl commands at both hours to confirm it, then raise the cache ratio before considering a bigger plan — a misconfigured cache is not fixed by extra cores.
Should a Spanish site have separate versions for Latin America?
Only when the pricing, the offer or the vocabulary genuinely differ. Start with one site in es-ES and hreflang declared. Each split doubles your certificates, backups, redirect chains and cache entries, and splitting early divides your authority across a distinction customers cannot yet see.
How heavy can a media page be before it fails the threshold?
There is no fixed byte limit; there is a 2.5 second LCP target at the 75th percentile of real visits. Trace the page, find the element being timed, then attack its export size, the font weights loading around it and any ad or analytics code executing before it paints.
Is Spanish consumer law a hosting problem?
No. Privacy notices and the legal notice are ordinary content under your control, SSL issues automatically on every hostname, and consent tooling is standard fitting. Your lawyer sets what has to be published; the platform simply presents it and does not get in the way.
Keep reading
Hosting in Spain
Rates in EUR, which domain endings we can register there, and how the tax works out.
Web Hosting in Paris
Paris on the same method, inside the same 10–35 ms European band.
Web Hosting in Warsaw
Warsaw, where the published band is 30–60 ms and the arithmetic shifts.
DNS Propagation (Glossary)
What propagation is, what the TTL controls, and how to watch a change land.
Node.js Hosting
Node.js with a real shell, for services a page cache cannot help.
Laravel Hosting
Laravel with Composer, SSH and Git deploys instead of uploading files.
Changing hosts? Run through our checklist first.
A straightforward sequence for a switch your visitors never feel: which files move first, how to shift email across without losing a single message, the right moment to repoint DNS, and the two mistakes behind almost all the downtime we get asked to rescue.
Test it at the hour that pays.
NVMe behind a server-level LiteSpeed cache from $2.42/mo, WebP built in, free SSL and an engineer-run migration at no charge.
View Node.js Hosting plans