Skip to content

Why Varnish

Varnish Cache is an open-source HTTP reverse proxy built for one job: keeping web content in memory and serving it as fast as the network allows. It has been in production for twenty years and sits in front of some of the most visited sites on the web. Every site on Edge Caching gets its own dedicated Varnish instance.

Varnish keeps complete HTTP responses in memory. Serving a cached page involves no PHP, no database query and no disk access: Varnish’s own processing time is measured in microseconds. What your visitors experience is essentially the network round trip, rather than the time WordPress needs to build the page.

A single Varnish server can answer tens of thousands of requests per second with modest CPU and memory. Traffic spikes from a newsletter, a news mention or a product launch are absorbed by the cache instead of reaching your server, where they would otherwise exhaust PHP workers and database connections.

Varnish’s behavior is defined in VCL, the Varnish Configuration Language, which is compiled to native code when it is loaded. VCL gives fine-grained control over every request and response: what is cached and for how long, which cookies are removed, which requests go straight to the origin, and which are rejected. On Edge Caching, each site’s VCL is generated from the settings you choose in the admin and reloaded without interrupting traffic.

Varnish can keep serving cached content after it has expired when the backend server stops responding, a feature called grace mode. Edge Caching builds Always Online on it, so your pages stay available while your server is down or under maintenance. See When your server is down.

Varnish is used by major platforms, news publishers, universities, open-source projects and content delivery networks. Fastly, one of the largest CDNs, is built on its own version of Varnish, so sites delivered by Fastly are served by Varnish technology as well.

These organizations served their pages through Varnish when checked on October 1, 2026, as shown by their public HTTP response headers:

Category Organization How it was verified
Platforms Wikipedia Wikimedia cache servers (x-cache: cp… hit), whose in-memory layer is Varnish, as described by Wikimedia engineers at SREcon EMEA 2025
Reddit Responses include Via: 1.1 varnish
Spotify Responses include Via: 1.1 varnish
News and media BBC Responses include Via: 1.1 varnish
CNN Responses include Via: 1.1 varnish
The New York Times Delivered through Fastly (X-Served-By: cache-…)
The Guardian Delivered through Fastly (X-Timer)
Forbes Responses include Via: 1.1 varnish
Vox Responses include Via: 1.1 varnish
BuzzFeed Responses include Via: 1.1 varnish
Le Monde Responses include Via: 1.1 varnish
Le Devoir Responses include Via: 1.1 varnish
Universities McGill University Responses include X-Varnish, from its own Varnish installation
University of Toronto Responses include Via: 1.1 varnish, delivered through Pantheon’s Fastly-based CDN
University of Ottawa Responses include Via: varnish
Open source Python.org Responses include Via: 1.1 varnish
PyPI, the Python Package Index Delivered through Fastly (X-Served-By: cache-…)
Drupal.org Responses include Via: 1.1 varnish

A site’s response headers usually reveal whether Varnish served it:

Terminal window
curl -sI https://www.example.com/ | grep -iE '^(via|x-varnish|x-cache|x-served-by|x-timer):'

Via: … varnish and X-Varnish indicate Varnish directly. X-Served-By: cache-… and X-Timer indicate Fastly’s Varnish-based platform. Sites served by Edge Caching identify themselves with X-Served-By: Layer 7 Edge Caching(TM), along with X-Cache: HIT or MISS.