{
  "model": "Qwen3.8-27B-FP8",
  "runtime_version": "0.26.0b2.dev1+g3b102b576",
  "aibrix_version": "v0.7.0",
  "gpu": "NVIDIA L20",
  "tp": 1,
  "max_model_len": 32768,
  "slo": {
    "ttft_seconds": 2,
    "tpot_seconds": 0.15,
    "e2e_seconds": 30
  },
  "quantile_method": "linear interpolation of successful complete streams; excludes rejected and failed requests",
  "notes": [
    "Synthetic performance load, not model quality evaluation.",
    "Admission proxy uses synthetic tenant headers; not native AIBrix tenant isolation.",
    "Cache phases preserve online engine and routing state; not a controlled cold-cache A/B.",
    "KPA wave stayed at two desired and Ready replicas; no scale event was demonstrated.",
    "Drain variants had different Pod placement; do not attribute throughput changes to shutdown settings."
  ],
  "phases": [
    {
      "phase": "fixed1-wave",
      "config": {
        "phase": "fixed1-wave",
        "repeat": 30,
        "output": 128,
        "route": "least-request",
        "inflight_cap": 48,
        "stages": [
          {
            "duration": 60,
            "rps": 0.2
          },
          {
            "duration": 240,
            "rps": 1.6
          },
          {
            "duration": 90,
            "rps": 0.2
          }
        ]
      },
      "elapsed": 392.345954015851,
      "requests": 309,
      "success": 309,
      "errors": 0,
      "dropped_arrivals": 105,
      "good": 28,
      "output_tokens_per_second": 100.80899164415004,
      "ttft": {
        "p50": 39.901676304638386,
        "p95": 40.38109262287617,
        "p99": 40.455603786706924
      },
      "tpot": {
        "p50": 0.061438812749592335,
        "p95": 0.06148603390755616,
        "p99": 0.06149819031475097
      },
      "e2e": {
        "p50": 47.70671582967043,
        "p95": 48.18672317266464,
        "p99": 48.26311762690544
      },
      "started_relative_seconds": 0.0,
      "finished_relative_seconds": 392.3467252254486,
      "tenants": {
        "A": {
          "requests": 309,
          "success": 309,
          "good": 28,
          "status_counts": {
            "200": 309
          },
          "input_tokens_range": [
            512,
            522
          ],
          "ttft_p95_seconds": 40.38109262287617,
          "tpot_p95_seconds": 0.06148603390755616
        }
      }
    },
    {
      "phase": "apa-cold-wave",
      "config": {
        "phase": "apa-cold-wave",
        "repeat": 30,
        "output": 128,
        "route": "least-request",
        "inflight_cap": 48,
        "stages": [
          {
            "duration": 60,
            "rps": 0.2
          },
          {
            "duration": 240,
            "rps": 1.6
          },
          {
            "duration": 90,
            "rps": 0.2
          }
        ]
      },
      "elapsed": 392.3328397125006,
      "requests": 309,
      "success": 309,
      "errors": 0,
      "dropped_arrivals": 105,
      "good": 28,
      "output_tokens_per_second": 100.81236133325851,
      "ttft": {
        "p50": 39.88237251341343,
        "p95": 40.398177574574945,
        "p99": 40.465438741147516
      },
      "tpot": {
        "p50": 0.0614209229669233,
        "p95": 0.06148465523334939,
        "p99": 0.061499916552089334
      },
      "e2e": {
        "p50": 47.69075945019722,
        "p95": 48.20229391157627,
        "p99": 48.27388119995594
      },
      "started_relative_seconds": 932.4455032348633,
      "finished_relative_seconds": 1324.7788314819336,
      "tenants": {
        "A": {
          "requests": 309,
          "success": 309,
          "good": 28,
          "status_counts": {
            "200": 309
          },
          "input_tokens_range": [
            512,
            522
          ],
          "ttft_p95_seconds": 40.398177574574945,
          "tpot_p95_seconds": 0.06148465523334939
        }
      }
    },
    {
      "phase": "apa-min2-wave",
      "config": {
        "phase": "apa-min2-wave",
        "repeat": 30,
        "output": 128,
        "route": "least-request",
        "inflight_cap": 48,
        "stages": [
          {
            "duration": 60,
            "rps": 0.2
          },
          {
            "duration": 240,
            "rps": 1.6
          },
          {
            "duration": 90,
            "rps": 0.2
          }
        ]
      },
      "elapsed": 392.3894816637039,
      "requests": 414,
      "success": 414,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 414,
      "output_tokens_per_second": 135.04949157994153,
      "ttft": {
        "p50": 0.2726028822362423,
        "p95": 0.29837810099124906,
        "p99": 0.30076953627169134
      },
      "tpot": {
        "p50": 0.05865061341777561,
        "p95": 0.059664945694522596,
        "p99": 0.059744584516981454
      },
      "e2e": {
        "p50": 7.728795420378447,
        "p95": 7.861676347255707,
        "p99": 7.873628860488534
      },
      "started_relative_seconds": 1887.4088773727417,
      "finished_relative_seconds": 2279.7988002300262,
      "tenants": {
        "A": {
          "requests": 414,
          "success": 414,
          "good": 414,
          "status_counts": {
            "200": 414
          },
          "input_tokens_range": [
            513,
            522
          ],
          "ttft_p95_seconds": 0.29837810099124906,
          "tpot_p95_seconds": 0.059664945694522596
        }
      }
    },
    {
      "phase": "fixed2-wave",
      "config": {
        "phase": "fixed2-wave",
        "repeat": 30,
        "output": 128,
        "route": "least-request",
        "inflight_cap": 48,
        "stages": [
          {
            "duration": 60,
            "rps": 0.2
          },
          {
            "duration": 240,
            "rps": 1.6
          },
          {
            "duration": 90,
            "rps": 0.2
          }
        ]
      },
      "elapsed": 392.3895588964224,
      "requests": 414,
      "success": 414,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 414,
      "output_tokens_per_second": 135.04946499860387,
      "ttft": {
        "p50": 0.2710700184106827,
        "p95": 0.2986324720084667,
        "p99": 0.3008790207654238
      },
      "tpot": {
        "p50": 0.05886959000013944,
        "p95": 0.05961882105375838,
        "p99": 0.059693391928522604
      },
      "e2e": {
        "p50": 7.750721704214811,
        "p95": 7.857001676782965,
        "p99": 7.868180607184768
      },
      "started_relative_seconds": 3352.3762674331665,
      "finished_relative_seconds": 3744.766389131546,
      "tenants": {
        "A": {
          "requests": 414,
          "success": 414,
          "good": 414,
          "status_counts": {
            "200": 414
          },
          "input_tokens_range": [
            512,
            522
          ],
          "ttft_p95_seconds": 0.2986324720084667,
          "tpot_p95_seconds": 0.05961882105375838
        }
      }
    },
    {
      "phase": "drain-native",
      "config": {
        "phase": "drain-native",
        "concurrency": 4,
        "duration": 180,
        "repeat": 30,
        "output": 512
      },
      "elapsed": 203.7812102586031,
      "requests": 24,
      "success": 22,
      "errors": 2,
      "dropped_arrivals": 0,
      "good": 22,
      "output_tokens_per_second": 55.27496860827218,
      "ttft": {
        "p50": 0.3834165707230568,
        "p95": 0.5359837368130683,
        "p99": 0.5412607724964619
      },
      "tpot": {
        "p50": 0.05346801510324917,
        "p95": 0.05388982946421302,
        "p99": 0.05606501197395026
      },
      "e2e": {
        "p50": 27.732523139566183,
        "p95": 27.92247476056218,
        "p99": 28.89585443511605
      },
      "started_relative_seconds": 4342.315498828888,
      "finished_relative_seconds": 4546.097085237503,
      "tenants": {
        "A": {
          "requests": 24,
          "success": 22,
          "good": 22,
          "status_counts": {
            "200": 24
          },
          "input_tokens_range": [
            513,
            522
          ],
          "ttft_p95_seconds": 0.5359837368130683,
          "tpot_p95_seconds": 0.05388982946421302
        }
      }
    },
    {
      "phase": "drain-graceful",
      "config": {
        "phase": "drain-graceful",
        "concurrency": 4,
        "duration": 180,
        "repeat": 30,
        "output": 512
      },
      "elapsed": 194.7581042200327,
      "requests": 28,
      "success": 28,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 28,
      "output_tokens_per_second": 73.60926035613674,
      "ttft": {
        "p50": 0.5854735560715199,
        "p95": 0.6343849528580904,
        "p99": 0.6352452652156353
      },
      "tpot": {
        "p50": 0.05337843148004985,
        "p95": 0.05388749792603247,
        "p99": 0.05388914720667435
      },
      "e2e": {
        "p50": 27.88128402084112,
        "p95": 27.916022704914212,
        "p99": 27.9179232878983
      },
      "started_relative_seconds": 8865.672135591507,
      "finished_relative_seconds": 9060.430705785751,
      "tenants": {
        "A": {
          "requests": 28,
          "success": 28,
          "good": 28,
          "status_counts": {
            "200": 28
          },
          "input_tokens_range": [
            511,
            521
          ],
          "ttft_p95_seconds": 0.6343849528580904,
          "tpot_p95_seconds": 0.05388749792603247
        }
      }
    },
    {
      "phase": "tenants-baseline",
      "config": {
        "repeat": 30,
        "route": "least-request",
        "stages": [
          {
            "duration": 150,
            "rps": 1.4
          }
        ],
        "inflight_cap": 48,
        "tenants": {
          "A": {
            "repeat": 30,
            "output": 64
          },
          "B": {
            "repeat": 90,
            "output": 512
          }
        },
        "phase": "tenants-baseline"
      },
      "elapsed": 223.45737256109715,
      "requests": 172,
      "success": 172,
      "errors": 0,
      "dropped_arrivals": 38,
      "good": 15,
      "output_tokens_per_second": 219.6750075300312,
      "ttft": {
        "p50": 25.23710623383522,
        "p95": 43.053326283395286,
        "p99": 46.691030851230025
      },
      "tpot": {
        "p50": 0.061240333482172726,
        "p95": 0.06593119872822648,
        "p99": 0.07242433642347652
      },
      "e2e": {
        "p50": 39.585700672119856,
        "p95": 72.93417028151453,
        "p99": 75.70323713079095
      },
      "started_relative_seconds": 9304.224569797516,
      "finished_relative_seconds": 9527.682366371155,
      "tenants": {
        "A": {
          "requests": 87,
          "success": 87,
          "good": 15,
          "status_counts": {
            "200": 87
          },
          "input_tokens_range": [
            514,
            522
          ],
          "ttft_p95_seconds": 42.293959968537095,
          "tpot_p95_seconds": 0.06849143496818014
        },
        "B": {
          "requests": 85,
          "success": 85,
          "good": 0,
          "status_counts": {
            "200": 85
          },
          "input_tokens_range": [
            1472,
            1482
          ],
          "ttft_p95_seconds": 43.86137460321188,
          "tpot_p95_seconds": 0.06279165382312935
        }
      }
    },
    {
      "phase": "tenants-protected",
      "config": {
        "repeat": 30,
        "route": "least-request",
        "stages": [
          {
            "duration": 150,
            "rps": 1.4
          }
        ],
        "inflight_cap": 48,
        "tenants": {
          "A": {
            "repeat": 30,
            "output": 64
          },
          "B": {
            "repeat": 90,
            "output": 512
          }
        },
        "phase": "tenants-protected"
      },
      "elapsed": 152.2747110053897,
      "requests": 210,
      "success": 115,
      "errors": 95,
      "dropped_arrivals": 0,
      "good": 115,
      "output_tokens_per_second": 77.75421093776319,
      "ttft": {
        "p50": 0.27838539332151413,
        "p95": 0.5574404694139957,
        "p99": 0.5880263376235962
      },
      "tpot": {
        "p50": 0.05570503229659701,
        "p95": 0.05782266118460231,
        "p99": 0.06687534065000596
      },
      "e2e": {
        "p50": 3.794935941696167,
        "p95": 28.486618676781653,
        "p99": 28.89825769469142
      },
      "started_relative_seconds": 9538.32793211937,
      "finished_relative_seconds": 9690.603073596954,
      "tenants": {
        "A": {
          "requests": 105,
          "success": 105,
          "good": 105,
          "status_counts": {
            "200": 105
          },
          "input_tokens_range": [
            513,
            522
          ],
          "ttft_p95_seconds": 0.30286300033330915,
          "tpot_p95_seconds": 0.0578649681238901
        },
        "B": {
          "requests": 105,
          "success": 10,
          "good": 10,
          "status_counts": {
            "429": 95,
            "200": 10
          },
          "input_tokens_range": [
            1477,
            1480
          ],
          "ttft_p95_seconds": 0.5915707375854253,
          "tpot_p95_seconds": 0.05555752341848764
        }
      }
    },
    {
      "phase": "cache-online-least-a",
      "config": {
        "phase": "cache-online-least-a",
        "concurrency": 4,
        "duration": 60,
        "repeat": 300,
        "output": 64,
        "shared_prefix": true,
        "route": "least-request"
      },
      "elapsed": 62.692971378564835,
      "requests": 68,
      "success": 68,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 68,
      "output_tokens_per_second": 69.41767002429843,
      "ttft": {
        "p50": 0.2655413784086704,
        "p95": 1.1975286073982696,
        "p99": 1.6898900564759969
      },
      "tpot": {
        "p50": 0.05329248734882899,
        "p95": 0.0536011961894849,
        "p99": 0.053619248438922185
      },
      "e2e": {
        "p50": 3.598590638488531,
        "p95": 4.54417694583535,
        "p99": 5.0564700708538295
      },
      "started_relative_seconds": 9703.348162174225,
      "finished_relative_seconds": 9766.041545391083,
      "tenants": {
        "A": {
          "requests": 68,
          "success": 68,
          "good": 68,
          "status_counts": {
            "200": 68
          },
          "input_tokens_range": [
            4831,
            4842
          ],
          "ttft_p95_seconds": 1.1975286073982696,
          "tpot_p95_seconds": 0.0536011961894849
        }
      }
    },
    {
      "phase": "cache-online-prefix-a",
      "config": {
        "phase": "cache-online-prefix-a",
        "concurrency": 4,
        "duration": 60,
        "repeat": 300,
        "output": 64,
        "shared_prefix": true,
        "route": "prefix-cache"
      },
      "elapsed": 61.46577537059784,
      "requests": 68,
      "success": 68,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 68,
      "output_tokens_per_second": 70.80362972337578,
      "ttft": {
        "p50": 0.21858476847410202,
        "p95": 0.28404670730233195,
        "p99": 0.2932561206817627
      },
      "tpot": {
        "p50": 0.05350490535298984,
        "p95": 0.05417660199815318,
        "p99": 0.05522300301917961
      },
      "e2e": {
        "p50": 3.6031925678253174,
        "p95": 3.614302673563361,
        "p99": 3.6610227274149656
      },
      "started_relative_seconds": 9776.114625930786,
      "finished_relative_seconds": 9837.580732584,
      "tenants": {
        "A": {
          "requests": 68,
          "success": 68,
          "good": 68,
          "status_counts": {
            "200": 68
          },
          "input_tokens_range": [
            4831,
            4842
          ],
          "ttft_p95_seconds": 0.28404670730233195,
          "tpot_p95_seconds": 0.05417660199815318
        }
      }
    },
    {
      "phase": "cache-online-prefix-b",
      "config": {
        "phase": "cache-online-prefix-b",
        "concurrency": 4,
        "duration": 60,
        "repeat": 300,
        "output": 64,
        "shared_prefix": true,
        "route": "prefix-cache"
      },
      "elapsed": 61.263291619718075,
      "requests": 68,
      "success": 68,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 68,
      "output_tokens_per_second": 71.03764562659043,
      "ttft": {
        "p50": 0.2599153704941273,
        "p95": 0.2909151565283537,
        "p99": 0.2945626311004162
      },
      "tpot": {
        "p50": 0.0530034349196487,
        "p95": 0.0535420951980447,
        "p99": 0.0535731813690019
      },
      "e2e": {
        "p50": 3.599989678710699,
        "p95": 3.6061027802526953,
        "p99": 3.625976641848683
      },
      "started_relative_seconds": 9848.634984731674,
      "finished_relative_seconds": 9909.898652791977,
      "tenants": {
        "A": {
          "requests": 68,
          "success": 68,
          "good": 68,
          "status_counts": {
            "200": 68
          },
          "input_tokens_range": [
            4834,
            4841
          ],
          "ttft_p95_seconds": 0.2909151565283537,
          "tpot_p95_seconds": 0.0535420951980447
        }
      }
    },
    {
      "phase": "cache-online-least-b",
      "config": {
        "phase": "cache-online-least-b",
        "concurrency": 4,
        "duration": 60,
        "repeat": 300,
        "output": 64,
        "shared_prefix": true,
        "route": "least-request"
      },
      "elapsed": 61.26087223738432,
      "requests": 68,
      "success": 68,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 68,
      "output_tokens_per_second": 71.04045112410596,
      "ttft": {
        "p50": 0.2488998882472515,
        "p95": 0.28729185834527016,
        "p99": 0.29063731551170346
      },
      "tpot": {
        "p50": 0.053010962075657315,
        "p95": 0.053547991356915896,
        "p99": 0.053596685560686246
      },
      "e2e": {
        "p50": 3.59902635589242,
        "p95": 3.606399295479059,
        "p99": 3.616635184213519
      },
      "started_relative_seconds": 9920.987095594406,
      "finished_relative_seconds": 9982.248330116272,
      "tenants": {
        "A": {
          "requests": 68,
          "success": 68,
          "good": 68,
          "status_counts": {
            "200": 68
          },
          "input_tokens_range": [
            4834,
            4842
          ],
          "ttft_p95_seconds": 0.28729185834527016,
          "tpot_p95_seconds": 0.053547991356915896
        }
      }
    },
    {
      "phase": "kpa-min2-wave",
      "config": {
        "phase": "kpa-min2-wave",
        "repeat": 30,
        "output": 128,
        "route": "least-request",
        "inflight_cap": 48,
        "stages": [
          {
            "duration": 60,
            "rps": 0.2
          },
          {
            "duration": 240,
            "rps": 1.6
          },
          {
            "duration": 90,
            "rps": 0.2
          }
        ]
      },
      "elapsed": 392.39199175685644,
      "requests": 414,
      "success": 414,
      "errors": 0,
      "dropped_arrivals": 0,
      "good": 414,
      "output_tokens_per_second": 135.04862768156644,
      "ttft": {
        "p50": 0.2700747177004814,
        "p95": 0.29862666986882686,
        "p99": 0.3021580930054188
      },
      "tpot": {
        "p50": 0.05885482512003794,
        "p95": 0.05975780751644157,
        "p99": 0.05985030486769095
      },
      "e2e": {
        "p50": 7.764125794172287,
        "p95": 7.869916563108563,
        "p99": 7.8868341839313505
      },
      "started_relative_seconds": 10123.869966983795,
      "finished_relative_seconds": 10516.26253604889,
      "tenants": {
        "A": {
          "requests": 414,
          "success": 414,
          "good": 414,
          "status_counts": {
            "200": 414
          },
          "input_tokens_range": [
            513,
            522
          ],
          "ttft_p95_seconds": 0.29862666986882686,
          "tpot_p95_seconds": 0.05975780751644157
        }
      }
    }
  ]
}