diff --git a/test/integration/scheduler_perf/README.md b/test/integration/scheduler_perf/README.md
index 682c7afd991..261dd5e776f 100644
--- a/test/integration/scheduler_perf/README.md
+++ b/test/integration/scheduler_perf/README.md
@@ -44,9 +44,9 @@ be used to select workloads. It works like GitHub label filtering: the flag acce
 a comma-separated list of label names. Each label may have a `+` or `-` as prefix. Labels with
 `+` or no prefix must be set for a workload for it to be run. `-` means that the label must not
 be set. For example, this runs all performance benchmarks except those that are labeled
-as "fast":
+as "integration-test":
 ```shell
-make test-integration WHAT=./test/integration/scheduler_perf ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling -perf-scheduling-label-filter=performance,-fast"
+make test-integration WHAT=./test/integration/scheduler_perf ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling -perf-scheduling-label-filter=performance,-integration-test"
 ```
 
 Once the benchmark is finished, JSON file with metrics is available in the current directory (test/integration/scheduler_perf). Look for `BenchmarkPerfScheduling_benchmark_YYYY-MM-DDTHH:MM:SSZ.json`.
@@ -139,6 +139,8 @@ The test cases labeled as `short` are executed in pull-kubernetes-integration jo
 | pull-kubernetes-integration      | integration-test,short |
 | ci-benchmark-scheduler-perf      | performance            |
 
+See the comment on [./config/performance-config.yaml](./config/performance-config.yaml) for the details.
+
 ## Scheduling throughput thresholds
 
 Thresholds are used to capture scheduler performance regressions in a periodic ci-benchmark-scheduler-perf job. 
diff --git a/test/integration/scheduler_perf/config/performance-config.yaml b/test/integration/scheduler_perf/config/performance-config.yaml
index e13d2fe1925..cb7649a6e71 100644
--- a/test/integration/scheduler_perf/config/performance-config.yaml
+++ b/test/integration/scheduler_perf/config/performance-config.yaml
@@ -1,18 +1,18 @@
-# The following labels are used in this file:
-# - fast: short execution time, ideally less than 30 seconds
-# - integration-test: used to select workloads that
-#   run in ci-kubernetes-integration-master. Choosing those tests
-#   is a tradeoff between code coverage and overall runtime.
-# - short: used to select workloads that
-#   run in pull-kubernetes-integration.
-#   We should make each test case with short label very small,
-#   so that all tests with the label should take less than 5 min to complete.
-#   They can be run using --short=true flag.
-# - performance: used to select workloads that run
-#   in ci-benchmark-scheduler-perf. Such workloads
-#   must run long enough (ideally, longer than 10 seconds)
-#   to provide meaningful samples for the pod scheduling
-#   rate.
+# The following labels are used in this file. (listed in ascending order of the number of covered test cases)
+#
+# - integration-test: test cases to run as the integration test, usually to spot some issues in the scheduler implementation or scheduler-perf itself.
+# - performance: test cases to run in the performance test.
+# - short: supplemental label for the above two labels (must not used alone), which literally means short execution time test cases.
+#
+# Specifically, the CIs use labels like the following:
+# - `ci-kubernetes-integration-master` (`integration-test`): Test cases are chosen based on a tradeoff between code coverage and overall runtime. 
+# It basically covers all test cases but with their smallest workload. 
+# - `pull-kubernetes-integration` (`integration-test`,`short`): Test cases are chosen so that they should take less than total 5 min to complete.
+# - `ci-benchmark-scheduler-perf` (`performance`): Long enough test cases are chosen (ideally, longer than 10 seconds) 
+# to provide meaningful samples for the pod scheduling rate.
+#
+# Also, `performance`+`short` isn't used in the CIs, but it's used to test the performance test locally.
+# (Sometimes, the test cases with `integration-test` are too small to spot issues.)
 #
 # Combining "performance" and "fast" selects suitable workloads for a local
 # before/after comparisons with benchstat.
@@ -29,19 +29,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 1000
@@ -71,19 +71,19 @@
     namespace: sched-1
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 1
       measurePods: 4
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 100
       measurePods: 400
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 1000
@@ -108,19 +108,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 5000
@@ -148,13 +148,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
@@ -197,13 +197,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
@@ -244,13 +244,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
@@ -290,19 +290,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 5000
@@ -332,13 +332,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
@@ -374,19 +374,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 5000
@@ -421,12 +421,12 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       measurePods: 10
   - name: 15000Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 390
     params:
       initNodes: 15000
@@ -448,19 +448,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 5
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 500
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 5000
@@ -490,19 +490,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 10
       measurePods: 10
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 1000
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 5000
@@ -532,13 +532,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 10
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 1000
@@ -593,13 +593,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5
       initPods: 2
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 200
@@ -631,13 +631,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 20
       measurePods: 5
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 18
     params:
       initNodes: 500
@@ -667,13 +667,13 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 20
       measurePods: 5
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 18
     params:
       initNodes: 500
@@ -702,19 +702,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes/2InitPods
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 5
       initPods: 2
       measurePods: 10
   - name: 500Nodes/200InitPods
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPods: 200
       measurePods: 1000
   - name: 5000Nodes/200InitPods
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       initPods: 200
@@ -750,17 +750,17 @@
     collectMetrics: true
   workloads:
   - name: 10Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 10
       measurePods: 100
   - name: 1000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 1000
       measurePods: 1000
   - name: 5000Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 5000
       measurePods: 2000
@@ -796,14 +796,14 @@
     namespace: measure-ns-0
   workloads:
   - name: 10Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 10
       initPodsPerNamespace: 2
       initNamespaces: 2
       measurePods: 6
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPodsPerNamespace: 4
@@ -850,14 +850,14 @@
     namespace: measure-ns-0
   workloads:
   - name: 10Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 10
       initPodsPerNamespace: 2
       initNamespaces: 2
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPodsPerNamespace: 4
@@ -907,14 +907,14 @@
     namespace: measure-ns-0
   workloads:
   - name: 10Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 10
       initPodsPerNamespace: 2
       initNamespaces: 2
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPodsPerNamespace: 4
@@ -961,14 +961,14 @@
     namespace: measure-ns-0
   workloads:
   - name: 10Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       initNodes: 10
       initPodsPerNamespace: 2
       initNamespaces: 2
       measurePods: 10
   - name: 500Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 500
       initPodsPerNamespace: 4
@@ -1005,19 +1005,19 @@
     collectMetrics: true
   workloads:
   - name: 5Nodes
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       taintNodes: 1
       normalNodes: 4
       measurePods: 4
   - name: 500Nodes
-    labels: [fast]
+    labels: [performance, short]
     params:
       taintNodes: 100
       normalNodes: 400
       measurePods: 400
   - name: 5000Nodes
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 68
     params:
       taintNodes: 1000
@@ -1060,7 +1060,7 @@
     collectMetrics: true
   workloads:
   - name: fast
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       # This testcase runs through all code paths without
       # taking too long overall.
@@ -1137,7 +1137,7 @@
     collectMetrics: true
   workloads:
   - name: fast
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       # This testcase runs through all code paths without
       # taking too long overall.
@@ -1203,7 +1203,7 @@
     collectMetrics: true
   workloads:
   - name: fast
-    labels: [integration-test, fast, short]
+    labels: [integration-test, performance, short]
     params:
       # This testcase runs through all code paths without
       # taking too long overall.
@@ -1275,7 +1275,7 @@
     collectMetrics: true
   workloads:
   - name: fast
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       # This testcase runs through all code paths without
       # taking too long overall.
@@ -1339,13 +1339,13 @@
     collectMetrics: true
   workloads:
   - name: 1Node_10GatedPods
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       gatedPods: 10
       deletingPods: 10
       measurePods: 10
   - name: 1Node_10000GatedPods
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 130
     params:
       gatedPods: 10000
@@ -1373,12 +1373,12 @@
     collectMetrics: true
   workloads:
   - name: 1Node_10GatedPods
-    labels: [fast]
+    labels: [performance, short]
     params:
       gatedPods: 10
       measurePods: 10
   - name: 1Node_10000GatedPods
-    labels: [performance, fast]
+    labels: [performance, short]
     threshold: 110
     params:
       gatedPods: 10000
@@ -1408,13 +1408,13 @@
     collectMetrics: true
   workloads:
   - name: 10Node_100DeletingPods
-    labels: [integration-test, fast]
+    labels: [integration-test, performance, short]
     params:
       initNodes: 10
       deletingPods: 10
       measurePods: 10
   - name: 1000Node_1000DeletingPods
-    labels: [performance, fast]
+    labels: [performance, short]
     params:
       initNodes: 1000
       deletingPods: 1000
diff --git a/test/integration/scheduler_perf/label_selector.go b/test/integration/scheduler_perf/label_selector.go
index 2d3a72bccdc..57f1c714db9 100644
--- a/test/integration/scheduler_perf/label_selector.go
+++ b/test/integration/scheduler_perf/label_selector.go
@@ -16,15 +16,13 @@ limitations under the License.
 
 package benchmark
 
-import "strings"
-
 // enabled checks a a label filter that works as in GitHub:
 // - empty string means enabled
 // - individual labels are comma-separated
 // - [+]<label> means the workload must have that label
 // - -<label> means the workload must not have that label
-func enabled(labelFilter string, labels ...string) bool {
-	for _, label := range strings.Split(labelFilter, ",") {
+func enabled(labelFilters []string, labels ...string) bool {
+	for _, label := range labelFilters {
 		if label == "" {
 			continue
 		}
diff --git a/test/integration/scheduler_perf/label_selector_test.go b/test/integration/scheduler_perf/label_selector_test.go
index 901e72dd600..e9dd566ae6c 100644
--- a/test/integration/scheduler_perf/label_selector_test.go
+++ b/test/integration/scheduler_perf/label_selector_test.go
@@ -60,7 +60,7 @@ func TestLabelFilter(t *testing.T) {
 		t.Run(labelFilter, func(t *testing.T) {
 			for labels, expected := range labelResults {
 				t.Run(labels, func(t *testing.T) {
-					actual := enabled(labelFilter, strings.Split(labels, ",")...)
+					actual := enabled(strings.Split(labelFilter, ","), strings.Split(labels, ",")...)
 					if actual != expected {
 						t.Errorf("expected enabled to be %v, got %v", expected, actual)
 					}
diff --git a/test/integration/scheduler_perf/scheduler_perf.go b/test/integration/scheduler_perf/scheduler_perf.go
index 142c0787add..4a5f7963120 100644
--- a/test/integration/scheduler_perf/scheduler_perf.go
+++ b/test/integration/scheduler_perf/scheduler_perf.go
@@ -812,8 +812,6 @@ func initTestOutput(tb testing.TB) io.Writer {
 	return output
 }
 
-var perfSchedulingLabelFilter = flag.String("perf-scheduling-label-filter", "performance", "comma-separated list of labels which a testcase must have (no prefix or +) or must not have (-), used by BenchmarkPerfScheduling")
-
 var specialFilenameChars = regexp.MustCompile(`[^a-zA-Z0-9-_]`)
 
 func setupTestCase(t testing.TB, tc *testCase, output io.Writer, outOfTreePluginRegistry frameworkruntime.Registry) (informers.SharedInformerFactory, ktesting.TContext) {
@@ -898,8 +896,10 @@ func setupTestCase(t testing.TB, tc *testCase, output io.Writer, outOfTreePlugin
 // RunBenchmarkPerfScheduling runs the scheduler performance tests.
 //
 // You can pass your own scheduler plugins via outOfTreePluginRegistry.
-// Also, you may want to put your plugins in PluginNames variable in this package.
-func RunBenchmarkPerfScheduling(b *testing.B, outOfTreePluginRegistry frameworkruntime.Registry) {
+// Also, you may want to put your plugins in PluginNames variable in this package
+// to collect metrics for them.
+// testcaseLabelSelectors is available to select specific test cases to run with labels on them.
+func RunBenchmarkPerfScheduling(b *testing.B, outOfTreePluginRegistry frameworkruntime.Registry, testcaseLabelSelectors []string) {
 	testCases, err := getTestCases(configFile)
 	if err != nil {
 		b.Fatal(err)
@@ -923,8 +923,8 @@ func RunBenchmarkPerfScheduling(b *testing.B, outOfTreePluginRegistry frameworkr
 		b.Run(tc.Name, func(b *testing.B) {
 			for _, w := range tc.Workloads {
 				b.Run(w.Name, func(b *testing.B) {
-					if !enabled(*perfSchedulingLabelFilter, append(tc.Labels, w.Labels...)...) {
-						b.Skipf("disabled by label filter %q", *perfSchedulingLabelFilter)
+					if !enabled(testcaseLabelSelectors, append(tc.Labels, w.Labels...)...) {
+						b.Skipf("disabled by label filter %v", testcaseLabelSelectors)
 					}
 
 					informerFactory, tCtx := setupTestCase(b, tc, output, outOfTreePluginRegistry)
@@ -974,8 +974,6 @@ func RunBenchmarkPerfScheduling(b *testing.B, outOfTreePluginRegistry frameworkr
 	}
 }
 
-var testSchedulingLabelFilter = flag.String("test-scheduling-label-filter", "integration-test", "comma-separated list of labels which a testcase must have (no prefix or +) or must not have (-), used by TestScheduling")
-
 func loadSchedulerConfig(file string) (*config.KubeSchedulerConfiguration, error) {
 	data, err := os.ReadFile(file)
 	if err != nil {
diff --git a/test/integration/scheduler_perf/scheduler_perf_test.go b/test/integration/scheduler_perf/scheduler_perf_test.go
index 9212653ceb1..0a2e4964311 100644
--- a/test/integration/scheduler_perf/scheduler_perf_test.go
+++ b/test/integration/scheduler_perf/scheduler_perf_test.go
@@ -19,11 +19,19 @@ limitations under the License.
 package benchmark_test
 
 import (
+	"flag"
+	"strings"
 	"testing"
 
 	benchmark "k8s.io/kubernetes/test/integration/scheduler_perf"
 )
 
+var perfSchedulingLabelFilter = flag.String("perf-scheduling-label-filter", "performance", "comma-separated list of labels which a testcase must have (no prefix or +) or must not have (-), used by BenchmarkPerfScheduling")
+
 func BenchmarkPerfScheduling(b *testing.B) {
-	benchmark.RunBenchmarkPerfScheduling(b, nil)
+	if testing.Short() {
+		*perfSchedulingLabelFilter += ",+short"
+	}
+
+	benchmark.RunBenchmarkPerfScheduling(b, nil, strings.Split(*perfSchedulingLabelFilter, ","))
 }
diff --git a/test/integration/scheduler_perf/scheduler_test.go b/test/integration/scheduler_perf/scheduler_test.go
index 2142d009dfc..a1e0ab2f26e 100644
--- a/test/integration/scheduler_perf/scheduler_test.go
+++ b/test/integration/scheduler_perf/scheduler_test.go
@@ -17,12 +17,16 @@ limitations under the License.
 package benchmark
 
 import (
+	"flag"
+	"strings"
 	"testing"
 
 	"k8s.io/component-base/metrics/legacyregistry"
 	"k8s.io/kubernetes/pkg/features"
 )
 
+var testSchedulingLabelFilter = flag.String("test-scheduling-label-filter", "integration-test,-performance", "comma-separated list of labels which a testcase must have (no prefix or +) or must not have (-), used by TestScheduling")
+
 func TestScheduling(t *testing.T) {
 	testCases, err := getTestCases(configFile)
 	if err != nil {
@@ -40,7 +44,7 @@ func TestScheduling(t *testing.T) {
 		t.Run(tc.Name, func(t *testing.T) {
 			for _, w := range tc.Workloads {
 				t.Run(w.Name, func(t *testing.T) {
-					if !enabled(*testSchedulingLabelFilter, append(tc.Labels, w.Labels...)...) {
+					if !enabled(strings.Split(*testSchedulingLabelFilter, ","), append(tc.Labels, w.Labels...)...) {
 						t.Skipf("disabled by label filter %q", *testSchedulingLabelFilter)
 					}
 					informerFactory, tCtx := setupTestCase(t, tc, nil, nil)