DRA: enable performance tracking with scheduler_perf

The performance of the basic "fill up the cluster"
scenario (SchedulingWithResourceClaimTemplate) and the steady-state
scenario (SteadyStateClusterResourceClaimTemplate) are relevant. The large
configurations should run long enough to provide meaningful results.

Performance may be different with queueing hints enabled, so variants with that
get added for those large configurations.

Author: pohly

test/integration/scheduler_perf/dra/performance-config.yaml

@@ -23,7 +23,6 @@
 - name: SchedulingWithResourceClaimTemplate
   featureGates:
     DynamicResourceAllocation: true
-    # SchedulerQueueingHints: true
   workloadTemplate:
   - opcode: createNodes
     countParam: $nodesWithoutDRA
@@ -91,6 +90,19 @@
       measurePods: 1000
       maxClaimsPerNode: 10
   - name: 5000pods_500nodes
+    featureGates:
+      SchedulerQueueingHints: false
+    labels: [performance]
+    params:
+      nodesWithDRA: 500
+      nodesWithoutDRA: 0
+      initPods: 2500
+      measurePods: 2500
+      maxClaimsPerNode: 10
+  - name: 5000pods_500nodes_QueueingHintsEnabled
+    featureGates:
+      SchedulerQueueingHints: true
+    labels: [performance]
     params:
       nodesWithDRA: 500
       nodesWithoutDRA: 0
@@ -108,7 +120,6 @@
 - name: SteadyStateClusterResourceClaimTemplate
   featureGates:
     DynamicResourceAllocation: true
-    # SchedulerQueueingHints: true
   workloadTemplate:
   - opcode: createNodes
     countParam: $nodesWithoutDRA
@@ -177,6 +188,19 @@
       maxClaimsPerNode: 10
       duration: 10s
   - name: empty_500nodes
+    featureGates:
+      SchedulerQueueingHints: false
+    labels: [performance]
+    params:
+      nodesWithDRA: 500
+      nodesWithoutDRA: 0
+      initClaims: 0
+      maxClaimsPerNode: 10
+      duration: 10s
+  - name: empty_500nodes_QueueingHintsEnabled
+    featureGates:
+      SchedulerQueueingHints: true
+    labels: [performance]
     params:
       nodesWithDRA: 500
       nodesWithoutDRA: 0
@@ -199,6 +223,19 @@
       maxClaimsPerNode: 10
       duration: 10s
   - name: half_500nodes
+    featureGates:
+      SchedulerQueueingHints: false
+    labels: [performance]
+    params:
+      nodesWithDRA: 500
+      nodesWithoutDRA: 0
+      initClaims: 2500
+      maxClaimsPerNode: 10
+      duration: 10s
+  - name: half_500nodes_QueueingHintsEnabled
+    featureGates:
+      SchedulerQueueingHints: true
+    labels: [performance]
     params:
       nodesWithDRA: 500
       nodesWithoutDRA: 0
@@ -221,6 +258,19 @@
       maxClaimsPerNode: 10
       duration: 10s
   - name: full_500nodes
+    featureGates:
+      SchedulerQueueingHints: false
+    labels: [performance]
+    params:
+      nodesWithDRA: 500
+      nodesWithoutDRA: 0
+      initClaims: 4990
+      maxClaimsPerNode: 10
+      duration: 10s
+  - name: full_500nodes_QueueingHintsEnabled
+    featureGates:
+      SchedulerQueueingHints: true
+    labels: [performance]
     params:
       nodesWithDRA: 500
       nodesWithoutDRA: 0