Merge pull request #134989 from alaypatel07/extended-resources-scheduler-perf

refactor scheduler_perf test for extended resources by dra to have 10-50 device classes

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

@@ -18,6 +18,7 @@
 # before/after comparisons with benchstat.
 
 # ExtendedResource uses pods with extended resources requests.
+# Each pod requests a unique extended resource.
 - name: ExtendedResource
   featureGates:
     DynamicResourceAllocation: true
@@ -58,17 +59,91 @@
       initPods: 0
       measurePods: 10
       maxClaimsPerNode: 10
+
+- name: ExtendedResource_25Classes
+  featureGates:
+    DynamicResourceAllocation: true
+    DRAExtendedResource: true
+  workloadTemplate:
+  - opcode: createNodes
+    countParam: $nodesWithoutDRA
+  - opcode: createNodes
+    nodeTemplatePath: ../templates/node-with-dra-test-driver.yaml
+    countParam: $nodesWithDRA
+  - opcode: createResourceDriver
+    driverName: test-driver.cdi.k8s.io
+    nodes: scheduler-perf-dra-*
+    maxClaimsPerNodeParam: $maxClaimsPerNode
+  - opcode: createAny
+    templatePath: ../templates/deviceclass-extended-resource.yaml
+    countParam: $classes
+  - opcode: createPods
+    namespace: init
+    countParam: $initPods
+    podTemplatePath: ../templates/pod-with-extended-resource-mod25.yaml
+  - opcode: createPods
+    namespace: test
+    countParam: $measurePods
+    podTemplatePath: ../templates/pod-with-extended-resource-mod25.yaml
+    collectMetrics: true
+  workloads:
   - name: 2000pods_200nodes
     params:
-      classes: 1000
+      classes: 25
       nodesWithDRA: 200
       nodesWithoutDRA: 0
       initPods: 1000
       measurePods: 1000
       maxClaimsPerNode: 10
   - name: 5000pods_500nodes
+    labels: [performance]
     params:
-      classes: 2500
+      classes: 25
+      nodesWithDRA: 500
+      nodesWithoutDRA: 0
+      initPods: 2500
+      measurePods: 2500
+      maxClaimsPerNode: 10
+
+- name: ExtendedResource_50Classes
+  featureGates:
+    DynamicResourceAllocation: true
+    DRAExtendedResource: true
+  workloadTemplate:
+  - opcode: createNodes
+    countParam: $nodesWithoutDRA
+  - opcode: createNodes
+    nodeTemplatePath: ../templates/node-with-dra-test-driver.yaml
+    countParam: $nodesWithDRA
+  - opcode: createResourceDriver
+    driverName: test-driver.cdi.k8s.io
+    nodes: scheduler-perf-dra-*
+    maxClaimsPerNodeParam: $maxClaimsPerNode
+  - opcode: createAny
+    templatePath: ../templates/deviceclass-extended-resource.yaml
+    countParam: $classes
+  - opcode: createPods
+    namespace: init
+    countParam: $initPods
+    podTemplatePath: ../templates/pod-with-extended-resource-mod50.yaml
+  - opcode: createPods
+    namespace: test
+    countParam: $measurePods
+    podTemplatePath: ../templates/pod-with-extended-resource-mod50.yaml
+    collectMetrics: true
+  workloads:
+  - name: 2000pods_200nodes
+    params:
+      classes: 50
+      nodesWithDRA: 200
+      nodesWithoutDRA: 0
+      initPods: 1000
+      measurePods: 1000
+      maxClaimsPerNode: 10
+  - name: 5000pods_500nodes
+    labels: [performance]
+    params:
+      classes: 50
       nodesWithDRA: 500
       nodesWithoutDRA: 0
       initPods: 2500

test/integration/scheduler_perf/dra/templates/pod-with-extended-resource-mod25.yaml

@@ -0,0 +1,16 @@
+# Pod template for extended resource tests.
+# Pods cycle through 25 device classes (example.com/gpu-0 through example.com/gpu-24)
+apiVersion: v1
+kind: Pod
+metadata:
+  name: test-dra-{{.Index}}
+spec:
+  containers:
+  - image: registry.k8s.io/pause:3.9
+    name: pause
+    resources:
+      requests:
+        example.com/gpu-{{mod .Index 25}}: 1
+      limits:
+        example.com/gpu-{{mod .Index 25}}: 1
+

test/integration/scheduler_perf/dra/templates/pod-with-extended-resource-mod50.yaml

@@ -0,0 +1,16 @@
+# Pod template for extended resource tests.
+# Pods cycle through 50 device classes (example.com/gpu-0 through example.com/gpu-49)
+apiVersion: v1
+kind: Pod
+metadata:
+  name: test-dra-{{.Index}}
+spec:
+  containers:
+  - image: registry.k8s.io/pause:3.9
+    name: pause
+    resources:
+      requests:
+        example.com/gpu-{{mod .Index 50}}: 1
+      limits:
+        example.com/gpu-{{mod .Index 50}}: 1
+