bench_concurrency

{“what”:“Surrogate primary key, auto-incrementing integer.”,“why”:“Uniquely identifies each row for joins and updates.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.064954847+00:00”}

{“what”:“UUID of the parent benchmark run.”,“why”:“Links all concurrency levels to the same run; without it levels are orphaned.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.088868900+00:00”}

{“what”:“Timestamp when this level’s metrics were recorded.”,“why”:“Enables time-based filtering and ordering of results.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.103336147+00:00”}

{“what”:“Name of the LLM provider (e.g., OpenAI, Anthropic).”,“why”:“Distinguishes results by vendor; without it cross-provider comparisons are impossible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.121245952+00:00”}

{“what”:“Model identifier as used by the provider.”,“why”:“Identifies which model produced the metrics; needed for per-model analysis.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.142297222+00:00”}

{“what”:“Concurrency level number (e.g., 1, 5, 10).”,“why”:“Defines the load level; without it results cannot be plotted against concurrency.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.191193691+00:00”}

{“what”:“Number of request waves executed at this level.”,“why”:“Indicates sample size; without it statistical confidence is unknown.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.205716010+00:00”}

{“what”:“Count of successful requests.”,“why”:“Base for success rate; without it success percentage is undefined.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.228785985+00:00”}

{“what”:“Count of requests rejected due to rate limiting.”,“why”:“Shows provider throttling; without it rate-limit impact is invisible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.242918357+00:00”}

{“what”:“Count of requests with 5xx server errors.”,“why”:“Identifies provider-side failures; without it error distribution is incomplete.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.257143196+00:00”}

{“what”:“Count of requests that timed out.”,“why”:“Reveals latency issues; without it timeout behavior is unknown.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.277219206+00:00”}

{“what”:“Count of requests failing for reasons not otherwise classified.”,“why”:“Captures residual errors; without it total error count is understated.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.293109909+00:00”}

{“what”:“Total completion tokens generated across all requests.”,“why”:“Measures output volume; without it token throughput cannot be computed.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.313382461+00:00”}

{“what”:“Aggregate token throughput in tokens per second.”,“why”:“Summarizes overall speed; without it performance comparison is impossible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.329473531+00:00”}

{“what”:“Per-stream token throughput in tokens per second, if measured.”,“why”:“Shows streaming efficiency; without it streaming performance is unknown.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.344173675+00:00”}

{“what”:“Median time-to-first-token in milliseconds.”,“why”:“Measures perceived latency; without it user experience is unquantified.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.362014330+00:00”}

{“what”:“99th percentile time-to-first-token in milliseconds.”,“why”:“Reveals tail latency; without it worst-case delays are hidden.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.379168593+00:00”}

{“what”:“Median total request duration in milliseconds.”,“why”:“Central tendency of end-to-end latency; without it typical response time is unknown.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.396301626+00:00”}

{“what”:“99th percentile total request duration in milliseconds.”,“why”:“Shows worst-case total latency; without it capacity planning is blind.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T00:01:52.416267650+00:00”}