vllm/examples/online_serving/dashboards/grafana/performance_statistics.json

{
  "annotations": {
    "list": [
      {
        "builtIn": 1,
        "datasource": {
          "type": "grafana",
          "uid": "-- Grafana --"
        },
        "enable": true,
        "hide": true,
        "iconColor": "rgba(0, 211, 255, 1)",
        "name": "Annotations & Alerts",
        "type": "dashboard"
      }
    ]
  },
  "editable": true,
  "fiscalYearStartMonth": 0,
  "graphTooltip": 0,
  "id": 26,
  "links": [],
  "panels": [
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 0
      },
      "id": 9,
      "panels": [],
      "title": "Graph: E2E latency over time ",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "End-to-End latency of requests, showing average and key percentiles over time.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisBorderShow": false,
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Latency",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "barWidthFactor": 0.6,
            "drawStyle": "line",
            "fillOpacity": 18,
            "gradientMode": "none",
            "hideFrom": {
              "legend": false,
              "tooltip": false,
              "viz": false
            },
            "insertNulls": false,
            "lineInterpolation": "linear",
            "lineWidth": 1,
            "pointSize": 5,
            "scaleDistribution": {
              "type": "linear"
            },
            "showPoints": "auto",
            "spanNulls": true,
            "stacking": {
              "group": "A",
              "mode": "none"
            },
            "thresholdsStyle": {
              "mode": "off"
            }
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 1
      },
      "id": 1,
      "options": {
        "legend": {
          "calcs": [],
          "displayMode": "list",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single",
          "sort": "none"
        }
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "rate(vllm:e2e_request_latency_seconds_sum[$__interval]) / rate(vllm:e2e_request_latency_seconds_count[$__interval])",
          "format": "table",
          "legendFormat": "E2E Latency",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "E2E Latency over Time",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "99th percentile of End-to-End request latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "P99",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 1
      },
      "id": 5,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:e2e_request_latency_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "E2E Latency (P99)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "90th percentile of End-to-End request latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "P90",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 1
      },
      "id": 4,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.90, sum by(le) (rate(vllm:e2e_request_latency_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "E2E Latency (P90)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average End-to-End request latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "Average",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 5
      },
      "id": 2,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "(sum(increase(vllm:e2e_request_latency_seconds_sum[$__range])) / sum(increase(vllm:e2e_request_latency_seconds_count[$__range])))",
          "legendFormat": "Average E2E Latency",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "E2E Latency (Avg)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "50th percentile (median) of End-to-End request latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "P50",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 5
      },
      "id": 3,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.50, sum by(le) (rate(vllm:e2e_request_latency_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "E2E Latency (P50)",
      "type": "stat"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 9
      },
      "id": 8,
      "panels": [],
      "title": "Graph: TTFT(Time To First Token) over time ",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Time to first token (TTFT) latency, showing average and key percentiles over time.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisBorderShow": false,
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Latency",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "barWidthFactor": 0.6,
            "drawStyle": "line",
            "fillOpacity": 18,
            "gradientMode": "none",
            "hideFrom": {
              "legend": false,
              "tooltip": false,
              "viz": false
            },
            "insertNulls": false,
            "lineInterpolation": "linear",
            "lineWidth": 1,
            "pointSize": 5,
            "scaleDistribution": {
              "type": "linear"
            },
            "showPoints": "auto",
            "spanNulls": false,
            "stacking": {
              "group": "A",
              "mode": "none"
            },
            "thresholdsStyle": {
              "mode": "off"
            }
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 10
      },
      "id": 10,
      "options": {
        "legend": {
          "calcs": [],
          "displayMode": "list",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single",
          "sort": "none"
        }
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "rate(vllm:time_to_first_token_seconds_sum[$__interval]) / rate(vllm:time_to_first_token_seconds_count[$__interval])",
          "format": "table",
          "legendFormat": "TTFT (Avg)",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "TTFT Over Time",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "99th percentile of Time To First Token latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "P99",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 10
      },
      "id": 14,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:time_to_first_token_seconds_bucket[$__range])))",
          "legendFormat": "TTFT (p99)",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "TTFT (P99)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "90th percentile of Time To First Token latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "P90",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 10
      },
      "id": 13,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.90, sum by(le) (rate(vllm:time_to_first_token_seconds_bucket[$__range])))",
          "legendFormat": "TTFT (p90)",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "TTFT (P90)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average Time To First Token latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "displayName": "Average",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 14
      },
      "id": 11,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "(sum(increase(vllm:time_to_first_token_seconds_sum[$__range])) / sum(increase(vllm:time_to_first_token_seconds_count[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "TTFT (Avg)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "50th percentile (median) of Time To First Token latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "displayName": "P50",
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 14
      },
      "id": 12,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orietitletChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.50, sum by(le) (rate(vllm:time_to_first_token_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "TTFT (P50)",
      "type": "stat"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 18
      },
      "id": 7,
      "panels": [],
      "title": "ITL (Iteration Latency / Time Per Output Token) over time.",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Iteration latency, or average time taken to generate a single output token, with percentiles.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisBorderShow": false,
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Latency",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "barWidthFactor": 0.6,
            "drawStyle": "line",
            "fillOpacity": 17,
            "gradientMode": "none",
            "hideFrom": {
              "legend": false,
              "tooltip": false,
              "viz": false
            },
            "insertNulls": false,
            "lineInterpolation": "linear",
            "lineWidth": 1,
            "pointSize": 5,
            "scaleDistribution": {
              "type": "linear"
            },
            "showPoints": "auto",
            "spanNulls": false,
            "stacking": {
              "group": "A",
              "mode": "none"
            },
            "thresholdsStyle": {
              "mode": "off"
            }
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 19
      },
      "id": 15,
      "options": {
        "legend": {
          "calcs": [],
          "displayMode": "list",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single",
          "sort": "none"
        }
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "rate(vllm:time_per_output_token_seconds_sum[$__interval]) / rate(vllm:time_per_output_token_seconds_count[$__interval])",
          "legendFormat": "ITL (Avg)",
          "range": true,
          "refId": "A"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "histogram_quantile(0.50, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__interval])))",
          "hide": false,
          "instant": false,
          "legendFormat": "ITL (p50)",
          "range": true,
          "refId": "B"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "histogram_quantile(0.90, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__interval])))",
          "hide": false,
          "instant": false,
          "legendFormat": "ITL (p90)",
          "range": true,
          "refId": "C"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__interval])))",
          "hide": false,
          "instant": false,
          "legendFormat": "ITL (p99)",
          "range": true,
          "refId": "D"
        }
      ],
      "title": "ITL (Time Per Output Token) Over Time",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "90th percentile of Iteration Latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 19
      },
      "id": 18,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.90, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "ITL (P90)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "99th percentile of Iteration Latency over the selected time range.\n\n",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 19
      },
      "id": 19,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "ITL (P99)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average Iteration Latency (time per output token) over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 23
      },
      "id": 16,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "(sum(increase(vllm:time_per_output_token_seconds_sum[$__range])) / sum(increase(vllm:time_per_output_token_seconds_count[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "ITL (Avg)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "50th percentile (median) of Iteration Latency over the selected time range.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "decimals": 2,
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 23
      },
      "id": 17,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "percentChangeColorMode": "standard",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showPercentChange": false,
        "textMode": "auto",
        "wideLayout": true
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "histogram_quantile(0.50, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket[$__range])))",
          "legendFormat": "__auto",
          "range": true,
          "refId": "A"
        }
      ],
      "title": "ITL (P50)",
      "type": "stat"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 27
      },
      "id": 6,
      "panels": [],
      "title": "TPS (Tokens Per Second)",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Rate of tokens processed per second, including prompt and generation phases.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisBorderShow": false,
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "barWidthFactor": 0.6,
            "drawStyle": "line",
            "fillOpacity": 0,
            "gradientMode": "none",
            "hideFrom": {
              "legend": false,
              "tooltip": false,
              "viz": false
            },
            "insertNulls": false,
            "lineInterpolation": "linear",
            "lineWidth": 1,
            "pointSize": 5,
            "scaleDistribution": {
              "type": "linear"
            },
            "showPoints": "auto",
            "spanNulls": false,
            "stacking": {
              "group": "A",
              "mode": "none"
            },
            "thresholdsStyle": {
              "mode": "off"
            }
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "tokens/sec (tps)"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 28
      },
      "id": 20,
      "options": {
        "legend": {
          "calcs": [],
          "displayMode": "list",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single",
          "sort": "none"
        }
      },
      "pluginVersion": "11.3.0",
      "targets": [
        {
          "editorMode": "code",
          "expr": "rate(vllm:generation_tokens_total[$__interval])",
          "legendFormat": "Generation TPS",
          "range": true,
          "refId": "A"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "rate(vllm:prompt_tokens_total[$__interval])",
          "hide": false,
          "instant": false,
          "legendFormat": "Prompt TPS",
          "range": true,
          "refId": "B"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "editorMode": "code",
          "expr": "rate(vllm:iteration_tokens_total_count[$__interval])",
          "hide": false,
          "instant": false,
          "legendFormat": "Overall Iteration TPS",
          "range": true,
          "refId": "C"
        }
      ],
      "title": "TPS (Tokens Per Second) Over Time",
      "type": "timeseries"
    }
  ],
  "preload": false,
  "schemaVersion": 40,
  "tags": [],
  "templating": {
    "list": [
      {
        "name": "DS_PROMETHEUS",
        "type": "datasource",
        "label": "datasource",
        "query": "prometheus",
        "refresh": 1,
        "current": {
          "text": "Prometheus",
          "value": "prometheus"
        }
      },
      {
        "current": {
          "text": "avg : Average\n0.50 : P50\n0.90 : P90\n0.99 : P99\n0.999 : Max (Approx)",
          "value": "avg : Average\n0.50 : P50\n0.90 : P90\n0.99 : P99\n0.999 : Max (Approx)"
        },
        "label": "Aggregation",
        "name": "agg_method",
        "options": [
          {
            "selected": true,
            "text": "avg : Average\n0.50 : P50\n0.90 : P90\n0.99 : P99\n0.999 : Max (Approx)",
            "value": "avg : Average\n0.50 : P50\n0.90 : P90\n0.99 : P99\n0.999 : Max (Approx)"
          }
        ],
        "query": "avg : Average\n0.50 : P50\n0.90 : P90\n0.99 : P99\n0.999 : Max (Approx)",
        "type": "custom"
      },
      {
        "current": {
          "text": [
            "granite-33-2b-instruct"
          ],
          "value": [
            "granite-33-2b-instruct"
          ]
        },
        "definition": "label_values(vllm:generation_tokens_total,model_name)",
        "includeAll": true,
        "label": "Deployment_ID",
        "multi": true,
        "name": "Deployment_id",
        "options": [],
        "query": {
          "qryType": 1,
          "query": "label_values(vllm:generation_tokens_total,model_name)",
          "refId": "PrometheusVariableQueryEditor-VariableQuery"
        },
        "refresh": 1,
        "regex": "",
        "type": "query"
      }
    ]
  },
  "time": {
    "from": "now-12h",
    "to": "now"
  },
  "timezone": "browser",
  "uid": "performance-statistics",
  "title": "Performance Statistics",
  "version": 40,
  "weekStart": ""
}