Jetty and Dropwizard Metrics: High-Throughput Timer Overhead

Dropwizard integrates the Metrics library with the Jetty handler chain to automatically track request-response cycles. This instrumentation relies on Timers to capture latency distributions and percentiles, which are essential for identifying performance bottlenecks in resource endpoints.

When monitoring extremely high-throughput, low-latency endpoints, there is a concern regarding the CPU and memory overhead introduced by high-precision timer collection. If the cost of measuring the request exceeds the execution time of the logic itself, the instrumentation may skew the resulting p99 latency data.

• Does the default Jetty integration in Dropwizard provide a mechanism to sample timer events rather than tracking every request?
• What is the expected performance impact on the Jetty thread pool when tracking a high volume of distinct resource timers?