I'm investigating intermittent performance issues on a large WooCommerce installation running on managed hosting.

Environment:

• 4 vCPU, 16 GB RAM
• MariaDB 10.11
• Redis Server 7.2
• Redis Object Cache 3.0.0 with Predis over a Unix socket
• Connection and read timeouts: 1 second

We experienced recurring Predis errors:

Error while reading line from the server

The plugin developer suggested testing 2.5 or 5-second timeouts.

During a hosting consultation, an experienced technician mentioned that Redis can require almost an entire CPU core under load and that the Redis server process continues running even after the WordPress cache plugin is disabled.

We've temporarily disabled Redis Object Cache, but WooCommerce still experiences occasional 10–18-second TTFB on dynamic requests.

My questions:

1. Is it correct that the Redis daemon normally continues running after WordPress object caching is disabled?
2. If no applications are actively using Redis, should its CPU consumption normally be negligible?
3. Under what workloads could Redis consume close to a full CPU core?
4. Which server-side metrics would you inspect to distinguish Redis CPU contention from PHP-FPM worker exhaustion or MariaDB bottlenecks?

We don't yet know whether Redis shares the same CPU allocation as PHP-FPM.

I'm interested in real-world operational experience rather than simply increasing server resources.

Thanks!