Building a Scalable Resiliant WP Setup

This my first post so Hi all,

I’m not new to IT and Linux but I’m new to running WP at scale and was hoping for some advice or insights.

I have customer who has a business that is starting to scale quite quickly and its taken us by surprise, I am using WP as a landing page for their marketing ads and to collect customer data, we have peak times for high load. I also need to build redundancy and resiliance in to the system as an outage at peak time or data loss=$$$.

At the moment I have a production server and a cold standby but they’re growing. I’m currently running a “traditional” linux VM, Apache, mariadb, postfix for a mailer.

I’ve looked at;

Linux VM: 2x HA proxy > 3x Apache/Wordpress/Mod_Security > 2x Master/Slave mysql, use of HyperDB plugin

Linux VM: 2x HA proxy > 3x Apache/Wordpress/Mod_Security > 2x Multimaster mysql, use of HyperDB plugin

Digital Ocean Droplets: 2x HA Proxy > 2x Apache/Wordpress/Mod_Security > Mysql & GlusterFS replicas

Similar to above but with OpenShift, Aws.

Advantages of Digital Ocean, AWS etc is the ability to be able to quickly rescale and add new services, agility, but have not run services like this in these before.

Any of the above but with Cloudflare to reduce the overhead of handling security and load balancing.

Interested to hear some of your experiences or insights,

Cheers,

Ady

Scaling with hardware alone is expensive. We’ve worked with a few high traffic sites and a lot of demanding ecommerce/WooCommerce sites. The way we’ve scaled those is by testing the code, optimising database calls, optimising the database itself and implemented full page caching.

Here’s a few tips to get you started

1. Use nginx instead of apache for better performance.
2. Use MariaDB instad of mysql.
3. Replace any get_posts and WP_Query calls with optimised $wpdb queries.
4. Use various optimisation tools eg. gtmetrix, pagespeed or yslow to check for problems
5. Use mysql general logging and check the speed for each and every query. Install indexes where needed.
6. Once optimisation is complete, implement some form of full page caching for pages that are not dynamic. Consider using ajax for dynamic elements.

If optimisation is not enough then you might want to look into load balancing or a cluster setup.