Skip to content

GitLab

Unverified Commit b39382ba authored by Ziqi Fan's avatar Ziqi Fan Committed by GitHub
Browse files

feat: add initial batch of KVBM metrics on match, offload and onboard (#2673)

parent 35055c6f
Hide whitespace changes
Inline Side-by-side
Showing with 809 additions and 214 deletions
deploy/metrics/grafana_dashboards/grafana-kvbm-dashboard.json
View file @ b39382ba
{
"annotations": {
"list": [
{
"builtIn": 1,
"datasource": {
"type": "grafana",
"uid": "-- Grafana --"
},
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
"name": "Annotations & Alerts",
"type": "dashboard"
}
]
},
"description": "All KVBM related metrics",
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 0,
"id": 4,
"links": [],
"panels": [
"annotations": {
"list": [
{
"builtIn": 1,
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
"type": "grafana",
"uid": "-- Grafana --"
},
"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"
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
"name": "Annotations & Alerts",
"type": "dashboard"
}
]
},
"description": "All KVBM related metrics",
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 0,
"id": 6,
"links": [],
"panels": [
{
"collapsed": false,
"gridPos": {
"h": 1,
"w": 24,
"x": 0,
"y": 0
},
"id": 7,
"panels": [],
"title": "General",
"type": "row"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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"
},
"thresholdsStyle": {
"mode": "off"
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 1
},
"id": 10,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_matched_tokens{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Matched Tokens",
"type": "timeseries"
},
{
"collapsed": false,
"gridPos": {
"h": 1,
"w": 24,
"x": 0,
"y": 9
},
"id": 5,
"panels": [],
"title": "Offload",
"type": "row"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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"
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "red",
"value": 80
}
]
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"overrides": []
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "red",
"value": 80
}
]
}
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 0
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 10
},
"id": 2,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"id": 1,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_offload_requests{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Offload Requests",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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"
}
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "red",
"value": 80
}
]
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_save_kv_layer_requests{dynamo_namespace=\"kvbm_connector_worker\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "KVBM Worker: save kv layer requests",
"type": "timeseries"
"overrides": []
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 10
},
"id": 3,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"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"
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_offload_blocks_d2h{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Offload Blocks",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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"
},
"thresholdsStyle": {
"mode": "off"
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 18
},
"id": 1,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_save_kv_layer_requests{dynamo_namespace=\"kvbm_connector_worker\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Save KV Layer Requests",
"type": "timeseries"
},
{
"collapsed": false,
"gridPos": {
"h": 1,
"w": 24,
"x": 0,
"y": 26
},
"id": 6,
"panels": [],
"title": "Onboard",
"type": "row"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "red",
"value": 80
}
]
"insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"overrides": []
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green"
},
{
"color": "red",
"value": 80
}
]
}
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 8
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 27
},
"id": 9,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"id": 2,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_onboard_requests{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Onboard Requests",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
"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"
},
{
"color": "red",
"value": 80
}
]
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_offload_requests{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "KVBM Leader: offload requests",
"type": "timeseries"
}
],
"preload": false,
"refresh": "auto",
"schemaVersion": 41,
"tags": [],
"templating": {
"list": []
},
"time": {
"from": "now-15m",
"to": "now"
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 27
},
"id": 4,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_onboard_blocks_h2d{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Onboard Blocks - Host to Device",
"type": "timeseries"
},
"timepicker": {},
"timezone": "browser",
"title": "KVBM Dashboard",
"uid": "3f679257-70a5-402c-92b4-05382337b548",
"version": 7
}
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"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"
},
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 35
},
"id": 8,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"hideZeros": false,
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "12.0.1",
"targets": [
{
"disableTextWrap": false,
"editorMode": "builder",
"expr": "dynamo_component_onboard_blocks_d2d{dynamo_namespace=\"kvbm_connector_leader\"}",
"fullMetaSearch": false,
"includeNullMetadata": true,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "Onboard Blocks - Disk to Host",
"type": "timeseries"
}
],
"preload": false,
"refresh": "auto",
"schemaVersion": 41,
"tags": [],
"templating": {
"list": []
},
"time": {
"from": "now-30m",
"to": "now"
},
"timepicker": {},
"timezone": "browser",
"title": "KVBM Dashboard",
"uid": "3f679257-70a5-402c-92b4-05382337b548",
"version": 5
}
\ No newline at end of file
lib/bindings/python/rust/llm/block_manager/vllm/connector/leader.rs
View file @ b39382ba
......@@ -80,6 +80,7 @@ pub struct KvConnectorLeader {
inflight_requests: HashSet<String>,
onboarding_slots: HashSet<String>,
iteration_counter: u64,
kvbm_metrics: KvbmMetrics,
}
impl KvConnectorLeader {
......@@ -114,12 +115,13 @@ impl KvConnectorLeader {
block_manager.clone(),
leader,
drt.clone(),
kvbm_metrics,
kvbm_metrics.clone(),
),
block_size,
inflight_requests: HashSet::new(),
onboarding_slots: HashSet::new(),
iteration_counter: 0,
kvbm_metrics,
}
}
}
......@@ -188,6 +190,9 @@ impl Leader for KvConnectorLeader {
"scheduling onboarding for {} external tokens",
num_external_tokens
);
self.kvbm_metrics
.matched_tokens
.inc_by(num_external_tokens as u64);
Ok((num_external_tokens, true))
} else {
Ok((0, false))
......
lib/bindings/python/rust/llm/block_manager/vllm/connector/leader/recorder.rs
View file @ b39382ba
......@@ -124,12 +124,13 @@ impl KvConnectorLeaderRecorder {
block_manager.clone(),
leader,
drt.clone(),
kvbm_metrics,
kvbm_metrics.clone(),
),
block_size,
inflight_requests: HashSet::new(),
onboarding_slots: HashSet::new(),
iteration_counter: 0,
kvbm_metrics,
};
let (unbounded_tx, unbounded_rx) = mpsc::unbounded_channel();
......
lib/bindings/python/rust/llm/block_manager/vllm/connector/leader/slot.rs
View file @ b39382ba
......@@ -197,7 +197,7 @@ impl<R: RequestKey> ConnectorSlotManager<R> {
let xfer_engine_task = CriticalTaskExecutionHandle::new_with_runtime(
|cancellation_token| async move {
xfer_engine
.execute(cancellation_token, drt_for_task, kvbm_metrics.clone())
.execute(cancellation_token, drt_for_task, kvbm_metrics)
.await
},
primary_token,
......@@ -1042,6 +1042,9 @@ impl LocalTransferEngine {
let leader_offload = Arc::clone(&self.leader);
let leader_onboard = Arc::clone(&self.leader);
let kvbm_metrics_onboard = kvbm_metrics.clone();
let kvbm_metrics_offload = kvbm_metrics.clone();
let onboard_task = CriticalTaskExecutionHandle::new_with_runtime(
|cancellation_token_onboard| async move {
while let Some(req) = onboard_rx.recv().await {
......@@ -1049,7 +1052,10 @@ impl LocalTransferEngine {
tracing::debug!("LocalOnboardTask: received cancellation signal");
break;
}
if let Err(e) = process_onboard_request(req, &leader_onboard).await {
if let Err(e) =
process_onboard_request(req, &leader_onboard, kvbm_metrics_onboard.clone())
.await
{
tracing::error!("LocalOnboardTask: error processing request: {:?}", e);
}
}
......@@ -1071,7 +1077,7 @@ impl LocalTransferEngine {
req,
&block_manager_offload,
&leader_offload,
kvbm_metrics.clone(),
kvbm_metrics_offload.clone(),
)
.await
{
......@@ -1145,6 +1151,9 @@ async fn process_offload_request(
kvbm_metrics: KvbmMetrics,
) -> anyhow::Result<()> {
kvbm_metrics.offload_requests.inc();
kvbm_metrics
.offload_blocks_d2h
.inc_by(offload_req.block_ids.len() as u64);
let request_id = &offload_req.request_id;
let operation_id = &offload_req.operation_id;
......@@ -1154,7 +1163,6 @@ async fn process_offload_request(
offload_req.block_ids.len()
);
// TODO: Implement actual offload logic
// 1. Acquire mutable host blocks
let host_blocks = block_manager
.host()
......@@ -1250,7 +1258,19 @@ async fn process_offload_request(
async fn process_onboard_request(
onboard_req: LocalOnboardRequest,
leader: &Arc<KvbmLeader>,
kvbm_metrics: KvbmMetrics,
) -> anyhow::Result<()> {
kvbm_metrics.onboard_requests.inc();
if onboard_req.src_blocks.storage_pool() == BlockTransferPool::Host {
kvbm_metrics
.onboard_blocks_h2d
.inc_by(onboard_req.src_blocks.len() as u64);
} else if onboard_req.src_blocks.storage_pool() == BlockTransferPool::Disk {
kvbm_metrics
.onboard_blocks_d2d
.inc_by(onboard_req.src_blocks.len() as u64);
}
let request_id = &onboard_req.request_id;
let operation_id = &onboard_req.operation_id;
......
lib/bindings/python/rust/llm/block_manager/vllm/connector/worker.rs
View file @ b39382ba
......@@ -265,7 +265,6 @@ impl Worker for KvConnectorWorker {
/// Trigger layer-wise completion signals.
/// Trigger block-wise completion signals afer last layer.
fn save_kv_layer(&mut self, _layer_name: String) -> anyhow::Result<()> {
self.kvbm_metrics.save_kv_layer_requests.inc();
self.layers_complete += 1;
if self.layers_complete == self.kv_cache_layers.len() {
let offloading_operations = std::mem::take(&mut self.offloading_operations);
......@@ -278,6 +277,7 @@ impl Worker for KvConnectorWorker {
self.connector.enqueue_request(operation);
}
}
self.kvbm_metrics.save_kv_layer_requests.inc();
Ok(())
}
......
lib/llm/src/block_manager/metrics_kvbm.rs
View file @ b39382ba
......@@ -6,8 +6,26 @@ use prometheus::IntCounter;
#[derive(Clone, Debug)]
pub struct KvbmMetrics {
// number of offload requests
pub offload_requests: IntCounter,
// number of blocks offloaded from device to host
pub offload_blocks_d2h: IntCounter,
// number of onboard requests
pub onboard_requests: IntCounter,
// number of blocks onboarded from host to device
pub onboard_blocks_h2d: IntCounter,
// number of blocks onboarded from disk to device
pub onboard_blocks_d2d: IntCounter,
// number of save kv layer requests
pub save_kv_layer_requests: IntCounter,
// number of matched tokens from KVBM
pub matched_tokens: IntCounter,
}
impl KvbmMetrics {
......@@ -15,6 +33,30 @@ impl KvbmMetrics {
let offload_requests = mr
.create_intcounter("offload_requests", "The number of offload requests", &[])
.unwrap();
let offload_blocks_d2h = mr
.create_intcounter(
"offload_blocks_d2h",
"The number of offload blocks from device to host",
&[],
)
.unwrap();
let onboard_requests = mr
.create_intcounter("onboard_requests", "The number of onboard requests", &[])
.unwrap();
let onboard_blocks_h2d = mr
.create_intcounter(
"onboard_blocks_h2d",
"The number of onboard blocks from host to device",
&[],
)
.unwrap();
let onboard_blocks_d2d = mr
.create_intcounter(
"onboard_blocks_d2d",
"The number of onboard blocks from disk to device",
&[],
)
.unwrap();
let save_kv_layer_requests = mr
.create_intcounter(
"save_kv_layer_requests",
......@@ -22,9 +64,17 @@ impl KvbmMetrics {
&[],
)
.unwrap();
let matched_tokens = mr
.create_intcounter("matched_tokens", "The number of matched tokens", &[])
.unwrap();
Self {
offload_requests,
offload_blocks_d2h,
onboard_requests,
onboard_blocks_h2d,
onboard_blocks_d2d,
save_kv_layer_requests,
matched_tokens,
}
}
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment