Request for Explanation and Best Practices for PCP (Performance Co-Pilot)

Issue

We are currently using Performance Co-Pilot (PCP) on a RHEL-based production system and would like to request detailed clarification and guidance regarding its components, usage, and best practices. Our key questions are:
- What is the recommended way to use PCP in a production environment?
- What are the main roles and interactions of the following components:

       pmcd
       pmlogger
       pmie
       Various pmdas (e.g., linux, proc, kvm, etc.)

• What is the function of pmie (Performance Metrics Inference Engine)? Is it essential to run in production, or is it optional?
• How do we properly configure thresholds or suppress noisy alerts generated by pmie?
• How can we manage log volume and disk usage efficiently when using PCP components?
• Are there recommendations for rotating or limiting PCP logs, especially under high-memory-demand conditions?
• Are there official performance tuning guides or templates for commonly used use-cases (e.g., monitoring JVM, databases, HAProxy)?
• We are running pmcd and pmlogger, and previously had pmie running until we disabled it due to high-frequency logs.