{ "execution_mode": "CPU simulation; arbitrary scenario coefficients, no serving engine executed", "experiment": "synthetic-arrival-and-goodput-scenarios", "scenario_assumptions": { "baseline": { "name": "baseline", "prefill_tokens_per_ms": 26.0, "prefix_cache_hit_efficiency": 0.45, "base_tpot_ms": 17.5, "batch_tpot_slope_ms": 0.62, "max_batch_slots": 16 }, "prefix_reuse": { "name": "prefix_reuse", "prefill_tokens_per_ms": 27.0, "prefix_cache_hit_efficiency": 0.92, "base_tpot_ms": 16.8, "batch_tpot_slope_ms": 0.58, "max_batch_slots": 16 }, "faster_decode": { "name": "faster_decode", "prefill_tokens_per_ms": 34.0, "prefix_cache_hit_efficiency": 0.5, "base_tpot_ms": 12.4, "batch_tpot_slope_ms": 0.44, "max_batch_slots": 16 } }, "slo_targets": { "ttft_slo_ms": 500.0, "tpot_slo_ms": 35.0 }, "trace_requests": 80, "scenario_comparisons": { "baseline": { "engine": "baseline", "workload": "agent-shared-prefix", "total_requests": 80, "duration_s": 22.912, "ttft_p50_ms": 3915.16, "ttft_p99_ms": 10235.3, "tpot_p50_ms": 27.42, "tpot_p99_ms": 27.42, "raw_output_tokens_per_s": 502.8, "request_throughput_rps": 3.492, "slo_goodput_rps": 0.567, "slo_attainment_ratio": 0.1625 }, "prefix_reuse": { "engine": "prefix_reuse", "workload": "agent-shared-prefix", "total_requests": 80, "duration_s": 21.224, "ttft_p50_ms": 3271.8, "ttft_p99_ms": 8821.9, "tpot_p50_ms": 26.08, "tpot_p99_ms": 26.08, "raw_output_tokens_per_s": 542.79, "request_throughput_rps": 3.769, "slo_goodput_rps": 0.754, "slo_attainment_ratio": 0.2 }, "faster_decode": { "engine": "faster_decode", "workload": "agent-shared-prefix", "total_requests": 80, "duration_s": 16.256, "ttft_p50_ms": 1781.48, "ttft_p99_ms": 5103.64, "tpot_p50_ms": 19.44, "tpot_p99_ms": 19.44, "raw_output_tokens_per_s": 708.66, "request_throughput_rps": 4.921, "slo_goodput_rps": 0.984, "slo_attainment_ratio": 0.2 } }, "diagnostics": { "overbatched_unbounded": { "engine": "overbatched_unbounded", "workload": "overbatched-tail-violation", "total_requests": 80, "duration_s": 15.035, "ttft_p50_ms": 283.89, "ttft_p99_ms": 2491.42, "tpot_p50_ms": 40.0, "tpot_p99_ms": 43.6, "raw_output_tokens_per_s": 766.23, "request_throughput_rps": 5.321, "slo_goodput_rps": 1.862, "slo_attainment_ratio": 0.35 }, "closed_loop_baseline": { "engine": "baseline", "workload": "closed-loop-illusion", "total_requests": 80, "duration_s": 213.48, "ttft_p50_ms": 80.29, "ttft_p99_ms": 87.68, "tpot_p50_ms": 18.12, "tpot_p99_ms": 18.12, "raw_output_tokens_per_s": 53.96, "request_throughput_rps": 0.375, "slo_goodput_rps": 0.375, "slo_attainment_ratio": 1.0 } } }