{
  "__inputs": [
    {
      "name": "DS_PROMETHEUS",
      "label": "prometheus",
      "description": "",
      "type": "datasource",
      "pluginId": "prometheus",
      "pluginName": "Prometheus"
    }
  ],
  "__elements": {},
  "__requires": [
    {
      "type": "grafana",
      "id": "grafana",
      "name": "Grafana",
      "version": "11.3.0"
    },
    {
      "type": "datasource",
      "id": "prometheus",
      "name": "Prometheus",
      "version": "1.0.0"
    },
    {
      "type": "panel",
      "id": "stat",
      "name": "Stat",
      "version": ""
    },
    {
      "type": "panel",
      "id": "timeseries",
      "name": "Time series",
      "version": ""
    }
  ],
  "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": 1,
  "id": null,
  "links": [],
  "liveNow": true,
  "panels": [
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 0
      },
      "id": 1,
      "panels": [],
      "title": "P/D Coordinator Overview",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average TTFT from vLLM prefill instances (prompt processing time)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              }
            ]
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 0,
        "y": 1
      },
      "id": 20,
      "options": {
        "colorMode": "value",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "lastNotNull"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(rate(vllm:time_to_first_token_seconds_sum{pod=~\".*prefill.*\"}[5m])) / sum(rate(vllm:time_to_first_token_seconds_count{pod=~\".*prefill.*\"}[5m])) * 1000",
          "refId": "A",
          "legendFormat": "Avg vLLM TTFT"
        }
      ],
      "title": "Avg vLLM TTFT (Prefill)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "True end-to-end TTFT from user perspective (vLLM Prefill TTFT + vLLM Decode TTFT)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "blue",
                "value": null
              }
            ]
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 8,
        "y": 1
      },
      "id": 24,
      "options": {
        "colorMode": "value",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "lastNotNull"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "(sum(rate(vllm:time_to_first_token_seconds_sum{pod=~\".*prefill.*\"}[5m])) / sum(rate(vllm:time_to_first_token_seconds_count{pod=~\".*prefill.*\"}[5m])) + sum(rate(vllm:time_to_first_token_seconds_sum{pod=~\".*decode.*\"}[5m])) / sum(rate(vllm:time_to_first_token_seconds_count{pod=~\".*decode.*\"}[5m]))) * 1000",
          "refId": "A",
          "legendFormat": "True End-to-End TTFT"
        }
      ],
      "title": "Avg True TTFT (Prefill + Decode)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average TTFT reported by vLLM decode instance (from Prometheus metrics)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "blue",
                "value": null
              }
            ]
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 16,
        "y": 1
      },
      "id": 21,
      "options": {
        "colorMode": "value",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "lastNotNull"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(rate(vllm:time_to_first_token_seconds_sum{pod=~\".*decode.*\"}[5m])) / sum(rate(vllm:time_to_first_token_seconds_count{pod=~\".*decode.*\"}[5m])) * 1000",
          "refId": "A",
          "legendFormat": "Avg vLLM TTFT"
        }
      ],
      "title": "Avg vLLM TTFT (Decode)",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Total number of P/D requests processed",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "blue",
                "value": null
              }
            ]
          },
          "unit": "short"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 0,
        "y": 31
      },
      "id": 23,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "count"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(vllm:e2e_request_latency_seconds_count{})",
          "refId": "A",
          "legendFormat": "Total Requests"
        }
      ],
      "title": "Total Requests",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "End-to-end request latency from vLLM including prefill and decode phases. Note: Does not include pd-proxy sidecar coordination overhead (~0-1ms).",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "E2E Latency (s)",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "drawStyle": "line",
            "fillOpacity": 20,
            "gradientMode": "none",
            "hideFrom": {
              "tooltip": false,
              "viz": false,
              "legend": false
            },
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "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": 10
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 5
      },
      "id": 2,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "multi",
          "sort": "none"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:e2e_request_latency_seconds_bucket{}[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:e2e_request_latency_seconds_bucket{}[$__rate_interval])))",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:e2e_request_latency_seconds_bucket{}[$__rate_interval])))",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "P/D E2E Latency (Prefill + Decode)",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Duration (s)",
            "axisPlacement": "auto",
            "barAlignment": 0,
            "drawStyle": "line",
            "fillOpacity": 20,
            "gradientMode": "none",
            "hideFrom": {
              "tooltip": false,
              "viz": false,
              "legend": false
            },
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "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
              }
            ]
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 5
      },
      "id": 3,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "multi",
          "sort": "none"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:e2e_request_latency_seconds_bucket{}[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:e2e_request_latency_seconds_bucket{}[$__rate_interval])))",
          "refId": "B",
          "legendFormat": "p95"
        }
      ],
      "title": "Total Request Duration (vLLM E2E)",
      "description": "End-to-end request latency from vLLM. Note: Does not include pd-proxy sidecar overhead (~0-1ms).",
      "type": "timeseries"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 13
      },
      "id": 10,
      "panels": [],
      "title": "P/D Component Breakdown",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Duration (s)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 14
      },
      "id": 4,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:request_prefill_time_seconds_bucket{}[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "p50"
        }
      ],
      "title": "vLLM Prefill Duration",
      "description": "Prefill phase duration from vLLM. Note: Does not include pd-proxy sidecar overhead (~0-1ms).",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Decode phase duration from vLLM. Note: Does not include pd-proxy sidecar overhead (~0-1ms).",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Duration (s)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 14
      },
      "id": 8,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:request_decode_time_seconds_bucket{}[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "p50"
        }
      ],
      "title": "vLLM Decode Duration",
      "type": "timeseries"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 22
      },
      "id": 11,
      "panels": [],
      "title": "Coordinator vs vLLM Metrics Comparison",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Prefill and decode phase durations from vLLM. Note: pd-proxy sidecar coordination overhead (~0-0.5ms) not shown as it requires pd-coordinator-metrics branch.",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Duration (s)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "s"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 23
      },
      "id": 6,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:request_prefill_time_seconds_bucket{}[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "vLLM Prefill (p50)"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:request_decode_time_seconds_bucket{}[$__rate_interval])))",
          "refId": "B",
          "legendFormat": "vLLM Decode (p50)"
        }
      ],
      "title": "P/D Phase Breakdown (vLLM View)",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "TTFT percentiles from vLLM decode instance (from Prometheus histogram)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "TTFT (ms)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 23
      },
      "id": 7,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:time_to_first_token_seconds_bucket{pod=~\".*decode.*\"}[5m]))) * 1000",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:time_to_first_token_seconds_bucket{pod=~\".*decode.*\"}[5m]))) * 1000",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:time_to_first_token_seconds_bucket{pod=~\".*decode.*\"}[5m]))) * 1000",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "vLLM TTFT (Decode Instance)",
      "type": "timeseries"
    },
    {
      "collapsed": false,
      "gridPos": {
        "h": 1,
        "w": 24,
        "x": 0,
        "y": 31
      },
      "id": 12,
      "panels": [],
      "title": "KV Cache Transfer Metrics (NIXL Connector)",
      "type": "row"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average KV cache transfer time (vLLM NIXL connector)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 50
              },
              {
                "color": "red",
                "value": 100
              }
            ]
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 0,
        "y": 32
      },
      "id": 13,
      "options": {
        "colorMode": "value",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "mean"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(rate(vllm:nixl_xfer_time_seconds_sum[$__rate_interval])) / sum(rate(vllm:nixl_xfer_time_seconds_count[$__rate_interval])) * 1000",
          "refId": "A",
          "legendFormat": "Avg Transfer Time",
          "instant": false
        }
      ],
      "title": "Avg KV Transfer Time",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Average bytes transferred per KV cache transfer",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "blue",
                "value": null
              }
            ]
          },
          "unit": "decmbytes"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 6,
        "y": 32
      },
      "id": 14,
      "options": {
        "colorMode": "value",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "mean"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(rate(vllm:nixl_bytes_transferred_sum[$__rate_interval])) / sum(rate(vllm:nixl_bytes_transferred_count[$__rate_interval])) / 1048576",
          "refId": "A",
          "legendFormat": "Avg MB Transferred",
          "instant": false
        }
      ],
      "title": "Avg MB per Transfer",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Total number of successful KV cache transfers",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "No Data",
                  "color": "text"
                }
              }
            }
          ],
          "noValue": "No Data",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              }
            ]
          },
          "unit": "short"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 32
      },
      "id": 15,
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "lastNotNull"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(increase(vllm:nixl_xfer_time_seconds_count[$__range]))",
          "refId": "A",
          "legendFormat": "Total Transfers",
          "instant": false
        }
      ],
      "title": "Total KV Transfers",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Total number of failed KV cache transfers",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "type": "special",
              "options": {
                "match": "null",
                "result": {
                  "text": "0",
                  "color": "green"
                }
              }
            }
          ],
          "noValue": "0",
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 1
              },
              {
                "color": "red",
                "value": 10
              }
            ]
          },
          "unit": "short"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 32
      },
      "id": 16,
      "options": {
        "colorMode": "background",
        "graphMode": "none",
        "justifyMode": "auto",
        "orientation": "auto",
        "reduceOptions": {
          "values": false,
          "calcs": [
            "lastNotNull"
          ],
          "fields": ""
        },
        "textMode": "auto"
      },
      "pluginVersion": "12.1.0",
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "sum(vllm:nixl_num_failed_transfers) or vector(0)",
          "refId": "A",
          "legendFormat": "Failed Transfers",
          "instant": false
        }
      ],
      "title": "Failed KV Transfers",
      "type": "stat"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "KV cache transfer time percentiles (p50, p95, p99)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Transfer Time (ms)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 36
      },
      "id": 17,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:nixl_xfer_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:nixl_xfer_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:nixl_xfer_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "KV Transfer Time Percentiles",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "KV cache transfer post time percentiles (time to complete post-transfer operations)",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Post Time (ms)",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "ms"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 36
      },
      "id": 18,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:nixl_post_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:nixl_post_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:nixl_post_time_seconds_bucket[$__rate_interval]))) * 1000",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "KV Transfer Post Time Percentiles",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Bytes transferred over time for KV cache transfers",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "MB Transferred",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "decmbytes"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 44
      },
      "id": 19,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:nixl_bytes_transferred_bucket[$__rate_interval]))) / 1048576",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:nixl_bytes_transferred_bucket[$__rate_interval]))) / 1048576",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:nixl_bytes_transferred_bucket[$__rate_interval]))) / 1048576",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "Bytes Transferred per Transfer",
      "type": "timeseries"
    },
    {
      "datasource": {
        "type": "prometheus",
        "uid": "${DS_PROMETHEUS}"
      },
      "description": "Number of NIXL descriptors used per transfer",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "palette-classic"
          },
          "custom": {
            "axisCenteredZero": false,
            "axisColorMode": "text",
            "axisLabel": "Descriptor Count",
            "axisPlacement": "auto",
            "drawStyle": "line",
            "fillOpacity": 20,
            "lineInterpolation": "smooth",
            "lineWidth": 2,
            "showPoints": "auto"
          },
          "unit": "short"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 44
      },
      "id": 20,
      "options": {
        "legend": {
          "calcs": [
            "mean",
            "max",
            "min"
          ],
          "displayMode": "table",
          "placement": "bottom",
          "showLegend": true
        },
        "tooltip": {
          "mode": "single"
        }
      },
      "targets": [
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.5, sum by (le) (rate(vllm:nixl_num_descriptors_bucket[$__rate_interval])))",
          "refId": "A",
          "legendFormat": "p50"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.95, sum by (le) (rate(vllm:nixl_num_descriptors_bucket[$__rate_interval])))",
          "refId": "B",
          "legendFormat": "p95"
        },
        {
          "datasource": {
            "type": "prometheus",
            "uid": "${DS_PROMETHEUS}"
          },
          "expr": "histogram_quantile(0.99, sum by (le) (rate(vllm:nixl_num_descriptors_bucket[$__rate_interval])))",
          "refId": "C",
          "legendFormat": "p99"
        }
      ],
      "title": "NIXL Descriptors per Transfer",
      "type": "timeseries"
    }
  ],
  "refresh": "30s",
  "schemaVersion": 38,
  "style": "dark",
  "tags": [
    "llm-d",
    "p/d",
    "tracing"
  ],
  "templating": {
    "list": []
  },
  "time": {
    "from": "now-1h",
    "to": "now"
  },
  "timepicker": {},
  "timezone": "",
  "title": "P/D Coordinator Metrics",
  "uid": "pd-coordinator-metrics",
  "version": 0,
  "weekStart": ""
}