How can you reduce AppDynamics agent overhead in a production environment?

Agent overhead in AppDynamics is generally low by design, but on very high-throughput or latency-sensitive services, a few concrete levers exist to trim it further without losing the visibility that matters most.

• Tune snapshot thresholds: if the slow/very-slow thresholds are too aggressive, the agent captures far more call graphs than needed; loosening them to match realistic SLAs cuts collection volume.
• Limit deep-dive instrumentation: disable or scope down method-level instrumentation (like info points or fine-grained custom exit point detection) on ultra-hot code paths where it isn't adding decision-relevant data.
• Reduce Business Transaction cardinality: excessive auto-discovered BTs (from unnormalized URLs) each carry their own baseline and metric bookkeeping; consolidating them with custom match rules reduces that overhead.
• Cap continuing-tier snapshot volume: rely on the built-in per-node snapshot rate limits rather than raising them unnecessarily during a broad incident.
• Right-size Analytics/log collection: if the Analytics Agent is capturing very high-cardinality transaction data, scope its collection to what's actually queried rather than everything by default.

The general principle is to measure before tuning: check the agent's own reported overhead metrics first, since guessing at optimizations without data often trims the wrong thing while leaving the actual cost driver untouched.