Adaptive Connection Pool Throttling based on Latency

In high-load environments, when a backend resource (like a JMS) starts slowing down, the standard behavior of a connection pool is to open more connections to handle the queue, up to the maxPoolSize. This often creates a "death spiral" where the increased number of concurrent connections further degrades the backend performance.

Currently, Liberty users must manually tune pools or rely on external circuit breakers. There is no native way to make the pool size sensitive to the health (latency) of the connection it provides.

Implement control monitor within the Liberty connection manager. This feature would monitor the average response time of requests using the pool.

Threshold Monitoring: Define a latency target threshold= 250ms (for example)

Dynamic Capping: If the actual average response time exceeds the threshold, the pool should dynamically throttle the effective maxPoolSize to prevent further congestion.

Recovery Phase: Once falls below the threshold for a sustained period (cool-down), the governor gradually restores the maxPoolSize to its original configured value.

Resilience: Automatically protects backends from being overwhelmed during latency spikes.

Stability: Prevents "cascading failures" across microservices.

Efficiency: Reduces the need for complex manual tuning and improves "out-of-the-box" stability for Liberty in unpredictable backend environments.