perinvocation

Perinvocation is a unit of measurement used to quantify each individual call to a function, service, or procedure. It treats every activation of a callable as a separate event, regardless of caller or context. This unit is commonly used in serverless computing, microservices, and API backends where many small, stateless executions occur.

In analytics and billing, per-invocation data is collected in logs and metrics to understand workload, latency, and cost. It enables comparison of the overhead of a single function call, monitoring throughput, and estimating scale and cost. Cloud providers often price per invocation, with additional charges tied to execution duration and resource allocation.

Interpreting per-invocation metrics can be affected by batching, retries, or asynchronous invocations. Cold starts, initialization overhead, and concurrency can distort average latency. Therefore, per-invocation measurements are typically complemented by latency percentiles (for example p95 or p99) and distribution analyses.

Using per-invocation metrics supports optimization strategies such as reducing call frequency, increasing caching, designing idempotent functions, and tuning performance. It is distinct from per-user or per-session metrics, which aggregate behavior across users or sessions rather than individual calls.