Provider Health

Provider health captures whether a provider should currently be considered eligible, preferred, or avoided for a given workload and region.

Health sources

Meridias derives health from request failures, latency degradation, timeout rates, and other operational anomalies observed in live traffic. These signals complement the slower-moving benchmark results.

Thresholds

Health states are driven by thresholds that determine when noise becomes a real operational concern. Threshold design should avoid both overreacting to isolated failures and masking sustained degradation.

Regional nuance

Provider performance can vary by geography, model, or workload segment. Health should therefore be tracked at the narrowest level that still gives the router usable signal.

Recovery

When a provider recovers, Meridias should reintroduce it cautiously so routing does not oscillate between healthy and degraded states. Controlled recovery protects production traffic from repeated flapping.

Next step

Review Failover Strategies for the actions taken when health drops below acceptable levels.