performance

We maintain our own Ubuntu servers located at Linode‘s London data centre, which allows us to optimise the configuration for performance. Piggybacking on the server infrastructure used for the Huddersfield Exposed site (est. 2015), the separate WordPress servers are capable of handling hundreds of requests per second with sub-millisecond response times.

The WordPress servers sit behind a front-end Nginx web server that delivers the static content and caches the dynamically-generated web pages for your visitors, ensuring pages load quickly. PHP FastCGI Process Manager (PHP-FPM) combined with PHP OPcache helps to keep each WordPress servers’ daily average CPU load at under 2%, leaving plenty of spare capacity to cope with sudden increases in visitors to your site.

If your site contains lots of images or multimedia, we can improve the performance by using Linode’s European cloud-based Object Storage CDN to deliver them.

We aim to achieve at least 99.9% uptime on the WordPress servers, which equates to an average downtime of 10 minutes per week.1

We also have an optional warm standby service which helps keep your site online during periods of planned and unplanned downtime.

When we recently benchmarked one of our WordPress sites, it was capable of delivering the site’s home page 5,145 times per second with a simulated load of 100 simultaneous users (equivalent to an average time per request of 0.194 milliseconds). When the load was increased to 1,000 simultaneous users, the server was still capable of maintaining a sub-millisecond response time (0.599ms), delivering the home page 1,670 times per second.


Notes:

  1. Planned downtime is usually limited to mandatory server reboots following patching. Reboots usually take around 2 minutes and are typically carried out at 2am. All of our servers are enrolled with the Ubuntu Canonical Livepatch Service.