diff --git a/pkg/client/typed/dynamic/client.go b/pkg/client/typed/dynamic/client.go
index dba2af0262..72c460b255 100644
--- a/pkg/client/typed/dynamic/client.go
+++ b/pkg/client/typed/dynamic/client.go
@@ -33,6 +33,7 @@ import (
 	"k8s.io/kubernetes/pkg/conversion/queryparams"
 	"k8s.io/kubernetes/pkg/runtime"
 	"k8s.io/kubernetes/pkg/runtime/serializer"
+	"k8s.io/kubernetes/pkg/util/flowcontrol"
 	"k8s.io/kubernetes/pkg/watch"
 )
 
@@ -78,6 +79,11 @@ func NewClient(conf *restclient.Config) (*Client, error) {
 	return &Client{cl: cl}, nil
 }
 
+// GetRateLimiter returns rate limier.
+func (c *Client) GetRateLimiter() flowcontrol.RateLimiter {
+	return c.cl.GetRateLimiter()
+}
+
 // Resource returns an API interface to the specified resource for this client's
 // group and version. If resource is not a namespaced resource, then namespace
 // is ignored.
diff --git a/pkg/controller/garbagecollector/garbagecollector.go b/pkg/controller/garbagecollector/garbagecollector.go
index 4cc4a854d4..3c922e04f9 100644
--- a/pkg/controller/garbagecollector/garbagecollector.go
+++ b/pkg/controller/garbagecollector/garbagecollector.go
@@ -36,6 +36,7 @@ import (
 	"k8s.io/kubernetes/pkg/controller/garbagecollector/metaonly"
 	"k8s.io/kubernetes/pkg/runtime"
 	"k8s.io/kubernetes/pkg/types"
+	"k8s.io/kubernetes/pkg/util/clock"
 	utilerrors "k8s.io/kubernetes/pkg/util/errors"
 	utilruntime "k8s.io/kubernetes/pkg/util/runtime"
 	"k8s.io/kubernetes/pkg/util/sets"
@@ -127,7 +128,7 @@ func (m *concurrentUIDToNode) Delete(uid types.UID) {
 }
 
 type Propagator struct {
-	eventQueue *workqueue.Type
+	eventQueue *workqueue.TimedWorkQueue
 	// uidToNode doesn't require a lock to protect, because only the
 	// single-threaded Propagator.processEvent() reads/writes it.
 	uidToNode *concurrentUIDToNode
@@ -311,7 +312,7 @@ func (gc *GarbageCollector) removeOrphanFinalizer(owner *node) error {
 // the "Orphan" finalizer. The node is add back into the orphanQueue if any of
 // these steps fail.
 func (gc *GarbageCollector) orphanFinalizer() {
-	key, quit := gc.orphanQueue.Get()
+	key, start, quit := gc.orphanQueue.Get()
 	if quit {
 		return
 	}
@@ -331,20 +332,21 @@ func (gc *GarbageCollector) orphanFinalizer() {
 	err := gc.orhpanDependents(owner.identity, dependents)
 	if err != nil {
 		glog.V(6).Infof("orphanDependents for %s failed with %v", owner.identity, err)
-		gc.orphanQueue.Add(owner)
+		gc.orphanQueue.AddWithTimestamp(owner, start)
 		return
 	}
 	// update the owner, remove "orphaningFinalizer" from its finalizers list
 	err = gc.removeOrphanFinalizer(owner)
 	if err != nil {
 		glog.V(6).Infof("removeOrphanFinalizer for %s failed with %v", owner.identity, err)
-		gc.orphanQueue.Add(owner)
+		gc.orphanQueue.AddWithTimestamp(owner, start)
 	}
+	OrphanProcessingLatency.Observe(sinceInMicroseconds(gc.clock, start))
 }
 
 // Dequeueing an event from eventQueue, updating graph, populating dirty_queue.
 func (p *Propagator) processEvent() {
-	key, quit := p.eventQueue.Get()
+	key, start, quit := p.eventQueue.Get()
 	if quit {
 		return
 	}
@@ -422,6 +424,7 @@ func (p *Propagator) processEvent() {
 			p.gc.dirtyQueue.Add(dep)
 		}
 	}
+	EventProcessingLatency.Observe(sinceInMicroseconds(p.gc.clock, start))
 }
 
 // GarbageCollector is responsible for carrying out cascading deletion, and
@@ -434,11 +437,14 @@ type GarbageCollector struct {
 	metaOnlyClientPool dynamic.ClientPool
 	// clientPool uses the regular dynamicCodec. We need it to update
 	// finalizers. It can be removed if we support patching finalizers.
-	clientPool  dynamic.ClientPool
-	dirtyQueue  *workqueue.Type
-	orphanQueue *workqueue.Type
-	monitors    []monitor
-	propagator  *Propagator
+	clientPool                       dynamic.ClientPool
+	dirtyQueue                       *workqueue.TimedWorkQueue
+	orphanQueue                      *workqueue.TimedWorkQueue
+	monitors                         []monitor
+	propagator                       *Propagator
+	clock                            clock.Clock
+	registeredRateLimiter            *RegisteredRateLimiter
+	registeredRateLimiterForMonitors *RegisteredRateLimiter
 }
 
 func gcListWatcher(client *dynamic.Client, resource unversioned.GroupVersionResource) *cache.ListWatch {
@@ -466,14 +472,15 @@ func gcListWatcher(client *dynamic.Client, resource unversioned.GroupVersionReso
 	}
 }
 
-func monitorFor(p *Propagator, clientPool dynamic.ClientPool, resource unversioned.GroupVersionResource, kind unversioned.GroupVersionKind) (monitor, error) {
+func (gc *GarbageCollector) monitorFor(resource unversioned.GroupVersionResource, kind unversioned.GroupVersionKind) (monitor, error) {
 	// TODO: consider store in one storage.
 	glog.V(6).Infof("create storage for resource %s", resource)
 	var monitor monitor
-	client, err := p.gc.metaOnlyClientPool.ClientForGroupVersion(resource.GroupVersion())
+	client, err := gc.metaOnlyClientPool.ClientForGroupVersion(resource.GroupVersion())
 	if err != nil {
 		return monitor, err
 	}
+	gc.registeredRateLimiterForMonitors.registerIfNotPresent(resource.GroupVersion(), client, "garbage_collector_monitoring")
 	setObjectTypeMeta := func(obj interface{}) {
 		runtimeObject, ok := obj.(runtime.Object)
 		if !ok {
@@ -493,13 +500,13 @@ func monitorFor(p *Propagator, clientPool dynamic.ClientPool, resource unversion
 					eventType: addEvent,
 					obj:       obj,
 				}
-				p.eventQueue.Add(event)
+				gc.propagator.eventQueue.Add(event)
 			},
 			UpdateFunc: func(oldObj, newObj interface{}) {
 				setObjectTypeMeta(newObj)
 				setObjectTypeMeta(oldObj)
 				event := event{updateEvent, newObj, oldObj}
-				p.eventQueue.Add(event)
+				gc.propagator.eventQueue.Add(event)
 			},
 			DeleteFunc: func(obj interface{}) {
 				// delta fifo may wrap the object in a cache.DeletedFinalStateUnknown, unwrap it
@@ -511,7 +518,7 @@ func monitorFor(p *Propagator, clientPool dynamic.ClientPool, resource unversion
 					eventType: deleteEvent,
 					obj:       obj,
 				}
-				p.eventQueue.Add(event)
+				gc.propagator.eventQueue.Add(event)
 			},
 		},
 	)
@@ -528,16 +535,20 @@ var ignoredResources = map[unversioned.GroupVersionResource]struct{}{
 }
 
 func NewGarbageCollector(metaOnlyClientPool dynamic.ClientPool, clientPool dynamic.ClientPool, resources []unversioned.GroupVersionResource) (*GarbageCollector, error) {
+	clock := clock.RealClock{}
 	gc := &GarbageCollector{
 		metaOnlyClientPool: metaOnlyClientPool,
 		clientPool:         clientPool,
-		dirtyQueue:         workqueue.New(),
-		orphanQueue:        workqueue.New(),
 		// TODO: should use a dynamic RESTMapper built from the discovery results.
-		restMapper: registered.RESTMapper(),
+		restMapper:                       registered.RESTMapper(),
+		clock:                            clock,
+		dirtyQueue:                       workqueue.NewTimedWorkQueue(clock),
+		orphanQueue:                      workqueue.NewTimedWorkQueue(clock),
+		registeredRateLimiter:            NewRegisteredRateLimiter(),
+		registeredRateLimiterForMonitors: NewRegisteredRateLimiter(),
 	}
 	gc.propagator = &Propagator{
-		eventQueue: workqueue.New(),
+		eventQueue: workqueue.NewTimedWorkQueue(gc.clock),
 		uidToNode: &concurrentUIDToNode{
 			RWMutex:   &sync.RWMutex{},
 			uidToNode: make(map[types.UID]*node),
@@ -553,7 +564,7 @@ func NewGarbageCollector(metaOnlyClientPool dynamic.ClientPool, clientPool dynam
 		if err != nil {
 			return nil, err
 		}
-		monitor, err := monitorFor(gc.propagator, gc.clientPool, resource, kind)
+		monitor, err := gc.monitorFor(resource, kind)
 		if err != nil {
 			return nil, err
 		}
@@ -563,7 +574,7 @@ func NewGarbageCollector(metaOnlyClientPool dynamic.ClientPool, clientPool dynam
 }
 
 func (gc *GarbageCollector) worker() {
-	key, quit := gc.dirtyQueue.Get()
+	key, start, quit := gc.dirtyQueue.Get()
 	if quit {
 		return
 	}
@@ -572,6 +583,7 @@ func (gc *GarbageCollector) worker() {
 	if err != nil {
 		utilruntime.HandleError(fmt.Errorf("Error syncing item %#v: %v", key, err))
 	}
+	DirtyProcessingLatency.Observe(sinceInMicroseconds(gc.clock, start))
 }
 
 // apiResource consults the REST mapper to translate an <apiVersion, kind,
@@ -594,6 +606,7 @@ func (gc *GarbageCollector) apiResource(apiVersion, kind string, namespaced bool
 func (gc *GarbageCollector) deleteObject(item objectReference) error {
 	fqKind := unversioned.FromAPIVersionAndKind(item.APIVersion, item.Kind)
 	client, err := gc.clientPool.ClientForGroupVersion(fqKind.GroupVersion())
+	gc.registeredRateLimiter.registerIfNotPresent(fqKind.GroupVersion(), client, "garbage_collector_operation")
 	resource, err := gc.apiResource(item.APIVersion, item.Kind, len(item.Namespace) != 0)
 	if err != nil {
 		return err
@@ -607,6 +620,7 @@ func (gc *GarbageCollector) deleteObject(item objectReference) error {
 func (gc *GarbageCollector) getObject(item objectReference) (*runtime.Unstructured, error) {
 	fqKind := unversioned.FromAPIVersionAndKind(item.APIVersion, item.Kind)
 	client, err := gc.clientPool.ClientForGroupVersion(fqKind.GroupVersion())
+	gc.registeredRateLimiter.registerIfNotPresent(fqKind.GroupVersion(), client, "garbage_collector_operation")
 	resource, err := gc.apiResource(item.APIVersion, item.Kind, len(item.Namespace) != 0)
 	if err != nil {
 		return nil, err
@@ -617,6 +631,7 @@ func (gc *GarbageCollector) getObject(item objectReference) (*runtime.Unstructur
 func (gc *GarbageCollector) updateObject(item objectReference, obj *runtime.Unstructured) (*runtime.Unstructured, error) {
 	fqKind := unversioned.FromAPIVersionAndKind(item.APIVersion, item.Kind)
 	client, err := gc.clientPool.ClientForGroupVersion(fqKind.GroupVersion())
+	gc.registeredRateLimiter.registerIfNotPresent(fqKind.GroupVersion(), client, "garbage_collector_operation")
 	resource, err := gc.apiResource(item.APIVersion, item.Kind, len(item.Namespace) != 0)
 	if err != nil {
 		return nil, err
@@ -627,6 +642,7 @@ func (gc *GarbageCollector) updateObject(item objectReference, obj *runtime.Unst
 func (gc *GarbageCollector) patchObject(item objectReference, patch []byte) (*runtime.Unstructured, error) {
 	fqKind := unversioned.FromAPIVersionAndKind(item.APIVersion, item.Kind)
 	client, err := gc.clientPool.ClientForGroupVersion(fqKind.GroupVersion())
+	gc.registeredRateLimiter.registerIfNotPresent(fqKind.GroupVersion(), client, "garbage_collector_operation")
 	resource, err := gc.apiResource(item.APIVersion, item.Kind, len(item.Namespace) != 0)
 	if err != nil {
 		return nil, err
@@ -752,6 +768,7 @@ func (gc *GarbageCollector) Run(workers int, stopCh <-chan struct{}) {
 		go wait.Until(gc.worker, 0, stopCh)
 		go wait.Until(gc.orphanFinalizer, 0, stopCh)
 	}
+	Register()
 	<-stopCh
 	glog.Infof("Garbage Collector: Shutting down")
 	gc.dirtyQueue.ShutDown()
diff --git a/pkg/controller/garbagecollector/garbagecollector_test.go b/pkg/controller/garbagecollector/garbagecollector_test.go
index 009fa5a562..90e748c359 100644
--- a/pkg/controller/garbagecollector/garbagecollector_test.go
+++ b/pkg/controller/garbagecollector/garbagecollector_test.go
@@ -35,6 +35,7 @@ import (
 	"k8s.io/kubernetes/pkg/client/restclient"
 	"k8s.io/kubernetes/pkg/client/typed/dynamic"
 	"k8s.io/kubernetes/pkg/types"
+	"k8s.io/kubernetes/pkg/util/clock"
 	"k8s.io/kubernetes/pkg/util/json"
 	"k8s.io/kubernetes/pkg/util/sets"
 	"k8s.io/kubernetes/pkg/util/workqueue"
@@ -282,13 +283,14 @@ func TestProcessEvent(t *testing.T) {
 
 	for _, scenario := range testScenarios {
 		propagator := &Propagator{
-			eventQueue: workqueue.New(),
+			eventQueue: workqueue.NewTimedWorkQueue(clock.RealClock{}),
 			uidToNode: &concurrentUIDToNode{
 				RWMutex:   &sync.RWMutex{},
 				uidToNode: make(map[types.UID]*node),
 			},
 			gc: &GarbageCollector{
-				dirtyQueue: workqueue.New(),
+				dirtyQueue: workqueue.NewTimedWorkQueue(clock.RealClock{}),
+				clock:      clock.RealClock{},
 			},
 		}
 		for i := 0; i < len(scenario.events); i++ {
diff --git a/pkg/controller/garbagecollector/metrics.go b/pkg/controller/garbagecollector/metrics.go
new file mode 100644
index 0000000000..af385e0d1e
--- /dev/null
+++ b/pkg/controller/garbagecollector/metrics.go
@@ -0,0 +1,73 @@
+/*
+Copyright 2016 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package garbagecollector
+
+import (
+	"sync"
+	"time"
+
+	"k8s.io/kubernetes/pkg/util/clock"
+
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+const (
+	GarbageCollectSubsystem    = "garbage_collector"
+	EventProcessingLatencyKey  = "event_processing_latency_microseconds"
+	DirtyProcessingLatencyKey  = "dirty_processing_latency_microseconds"
+	OrphanProcessingLatencyKey = "orphan_processing_latency_microseconds"
+)
+
+var (
+	EventProcessingLatency = prometheus.NewSummary(
+		prometheus.SummaryOpts{
+			Subsystem: GarbageCollectSubsystem,
+			Name:      EventProcessingLatencyKey,
+			Help:      "Time in microseconds of an event spend in the eventQueue",
+		},
+	)
+	DirtyProcessingLatency = prometheus.NewSummary(
+		prometheus.SummaryOpts{
+			Subsystem: GarbageCollectSubsystem,
+			Name:      DirtyProcessingLatencyKey,
+			Help:      "Time in microseconds of an item spend in the dirtyQueue",
+		},
+	)
+	OrphanProcessingLatency = prometheus.NewSummary(
+		prometheus.SummaryOpts{
+			Subsystem: GarbageCollectSubsystem,
+			Name:      OrphanProcessingLatencyKey,
+			Help:      "Time in microseconds of an item spend in the orphanQueue",
+		},
+	)
+)
+
+var registerMetrics sync.Once
+
+// Register all metrics.
+func Register() {
+	// Register the metrics.
+	registerMetrics.Do(func() {
+		prometheus.MustRegister(EventProcessingLatency)
+		prometheus.MustRegister(DirtyProcessingLatency)
+		prometheus.MustRegister(OrphanProcessingLatency)
+	})
+}
+
+func sinceInMicroseconds(clock clock.Clock, start time.Time) float64 {
+	return float64(clock.Since(start).Nanoseconds() / time.Microsecond.Nanoseconds())
+}
diff --git a/pkg/controller/garbagecollector/rate_limiter_helper.go b/pkg/controller/garbagecollector/rate_limiter_helper.go
new file mode 100644
index 0000000000..167d4251b0
--- /dev/null
+++ b/pkg/controller/garbagecollector/rate_limiter_helper.go
@@ -0,0 +1,59 @@
+/*
+Copyright 2016 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package garbagecollector
+
+import (
+	"fmt"
+	"strings"
+	"sync"
+
+	"k8s.io/kubernetes/pkg/api/unversioned"
+	"k8s.io/kubernetes/pkg/client/typed/dynamic"
+	"k8s.io/kubernetes/pkg/util/metrics"
+)
+
+// RegisteredRateLimiter records the registered RateLimters to avoid
+// duplication.
+type RegisteredRateLimiter struct {
+	rateLimiters map[unversioned.GroupVersion]struct{}
+	lock         sync.RWMutex
+}
+
+// NewRegisteredRateLimiter returns a new RegisteredRateLimiater.
+func NewRegisteredRateLimiter() *RegisteredRateLimiter {
+	return &RegisteredRateLimiter{
+		rateLimiters: make(map[unversioned.GroupVersion]struct{}),
+	}
+}
+
+func (r *RegisteredRateLimiter) registerIfNotPresent(gv unversioned.GroupVersion, client *dynamic.Client, prefix string) {
+	r.lock.RLock()
+	_, ok := r.rateLimiters[gv]
+	r.lock.RUnlock()
+	if ok {
+		return
+	}
+	r.lock.Lock()
+	defer r.lock.Unlock()
+	if _, ok := r.rateLimiters[gv]; !ok {
+		if rateLimiter := client.GetRateLimiter(); rateLimiter != nil {
+			group := strings.Replace(gv.Group, ".", ":", -1)
+			metrics.RegisterMetricAndTrackRateLimiterUsage(fmt.Sprintf("%s_%s_%s", prefix, group, gv.Version), rateLimiter)
+		}
+		r.rateLimiters[gv] = struct{}{}
+	}
+}
diff --git a/pkg/util/workqueue/timed_queue.go b/pkg/util/workqueue/timed_queue.go
new file mode 100644
index 0000000000..d22ed8bfca
--- /dev/null
+++ b/pkg/util/workqueue/timed_queue.go
@@ -0,0 +1,62 @@
+/*
+Copyright 2016 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package workqueue
+
+import (
+	"time"
+
+	"k8s.io/kubernetes/pkg/util/clock"
+)
+
+type TimedWorkQueue struct {
+	*Type
+	clock clock.Clock
+}
+
+type timedWorkQueueItem struct {
+	time time.Time
+	obj  interface{}
+}
+
+func NewTimedWorkQueue(clock clock.Clock) *TimedWorkQueue {
+	return &TimedWorkQueue{New(), clock}
+}
+
+// Add adds the obj along with the current timestamp to the queue.
+func (q TimedWorkQueue) Add(obj interface{}) {
+	start := q.clock.Now()
+	item := timedWorkQueueItem{start, obj}
+	q.Type.Add(item)
+}
+
+// AddWithTimestamp is useful if the caller does not want to refresh the start
+// time when requeuing an item.
+func (q TimedWorkQueue) AddWithTimestamp(obj interface{}, timestamp time.Time) {
+	item := timedWorkQueueItem{timestamp, obj}
+	q.Type.Add(item)
+}
+
+// Get gets the obj along with its timestamp from the queue.
+func (q TimedWorkQueue) Get() (item interface{}, start time.Time, shutdown bool) {
+	item, shutdown = q.Type.Get()
+	if item != nil {
+		timed, _ := item.(timedWorkQueueItem)
+		item = timed.obj
+		start = timed.time
+	}
+	return item, start, shutdown
+}
diff --git a/test/e2e/framework/metrics_util.go b/test/e2e/framework/metrics_util.go
index 753e8f2387..66a6d30ee4 100644
--- a/test/e2e/framework/metrics_util.go
+++ b/test/e2e/framework/metrics_util.go
@@ -57,6 +57,10 @@ func (m *MetricsForE2E) filterMetrics() {
 	for _, metric := range InterestingApiServerMetrics {
 		interestingApiServerMetrics[metric] = (*m).ApiServerMetrics[metric]
 	}
+	interestingControllerManagerMetrics := make(metrics.ControllerManagerMetrics)
+	for _, metric := range InterestingControllerManagerMetrics {
+		interestingControllerManagerMetrics[metric] = (*m).ControllerManagerMetrics[metric]
+	}
 	interestingKubeletMetrics := make(map[string]metrics.KubeletMetrics)
 	for kubelet, grabbed := range (*m).KubeletMetrics {
 		interestingKubeletMetrics[kubelet] = make(metrics.KubeletMetrics)
@@ -65,6 +69,7 @@ func (m *MetricsForE2E) filterMetrics() {
 		}
 	}
 	(*m).ApiServerMetrics = interestingApiServerMetrics
+	(*m).ControllerManagerMetrics = interestingControllerManagerMetrics
 	(*m).KubeletMetrics = interestingKubeletMetrics
 }
 
@@ -76,6 +81,12 @@ func (m *MetricsForE2E) PrintHumanReadable() string {
 			buf.WriteString(fmt.Sprintf("\t%v\n", metrics.PrintSample(sample)))
 		}
 	}
+	for _, interestingMetric := range InterestingControllerManagerMetrics {
+		buf.WriteString(fmt.Sprintf("For %v:\n", interestingMetric))
+		for _, sample := range (*m).ControllerManagerMetrics[interestingMetric] {
+			buf.WriteString(fmt.Sprintf("\t%v\n", metrics.PrintSample(sample)))
+		}
+	}
 	for kubelet, grabbed := range (*m).KubeletMetrics {
 		buf.WriteString(fmt.Sprintf("For %v:\n", kubelet))
 		for _, interestingMetric := range InterestingKubeletMetrics {
@@ -104,6 +115,12 @@ var InterestingApiServerMetrics = []string{
 	"etcd_request_latencies_summary",
 }
 
+var InterestingControllerManagerMetrics = []string{
+	"garbage_collector_event_processing_latency_microseconds",
+	"garbage_collector_dirty_processing_latency_microseconds",
+	"garbage_collector_orphan_processing_latency_microseconds",
+}
+
 var InterestingKubeletMetrics = []string{
 	"kubelet_container_manager_latency_microseconds",
 	"kubelet_docker_errors",
diff --git a/test/e2e/garbage_collector.go b/test/e2e/garbage_collector.go
index 43fe3e0965..de8e9eba7d 100644
--- a/test/e2e/garbage_collector.go
+++ b/test/e2e/garbage_collector.go
@@ -24,6 +24,7 @@ import (
 	"k8s.io/kubernetes/pkg/api/unversioned"
 	"k8s.io/kubernetes/pkg/api/v1"
 	clientset "k8s.io/kubernetes/pkg/client/clientset_generated/release_1_3"
+	"k8s.io/kubernetes/pkg/metrics"
 	"k8s.io/kubernetes/pkg/util/wait"
 	"k8s.io/kubernetes/test/e2e/framework"
 
@@ -97,6 +98,23 @@ func verifyRemainingObjects(f *framework.Framework, clientSet clientset.Interfac
 	return ret, nil
 }
 
+func gatherMetrics(f *framework.Framework) {
+	By("Gathering metrics")
+	var summary framework.TestDataSummary
+	grabber, err := metrics.NewMetricsGrabber(f.Client, false, false, true, false)
+	if err != nil {
+		framework.Logf("Failed to create MetricsGrabber. Skipping metrics gathering.")
+	} else {
+		received, err := grabber.Grab()
+		if err != nil {
+			framework.Logf("MetricsGrabber failed grab metrics. Skipping metrics gathering.")
+		} else {
+			summary = (*framework.MetricsForE2E)(&received)
+			framework.Logf(summary.PrintHumanReadable())
+		}
+	}
+}
+
 var _ = framework.KubeDescribe("Garbage collector", func() {
 	f := framework.NewDefaultFramework("gc")
 	It("[Feature:GarbageCollector] should delete pods created by rc when not orphaning", func() {
@@ -147,6 +165,7 @@ var _ = framework.KubeDescribe("Garbage collector", func() {
 				framework.Failf("remaining pods are: %#v", remainingPods)
 			}
 		}
+		gatherMetrics(f)
 	})
 
 	It("[Feature:GarbageCollector] should orphan pods created by rc", func() {
@@ -193,5 +212,6 @@ var _ = framework.KubeDescribe("Garbage collector", func() {
 		}); err != nil && err != wait.ErrWaitTimeout {
 			framework.Failf("%v", err)
 		}
+		gatherMetrics(f)
 	})
 })
diff --git a/test/integration/garbagecollector/garbage_collector_test.go b/test/integration/garbagecollector/garbage_collector_test.go
index d42cbede37..226d3482ad 100644
--- a/test/integration/garbagecollector/garbage_collector_test.go
+++ b/test/integration/garbagecollector/garbage_collector_test.go
@@ -27,6 +27,7 @@ import (
 	"testing"
 	"time"
 
+	dto "github.com/prometheus/client_model/go"
 	"k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/errors"
 	"k8s.io/kubernetes/pkg/api/unversioned"
@@ -424,6 +425,19 @@ func TestStressingCascadingDeletion(t *testing.T) {
 	if gc.GraphHasUID(uids) {
 		t.Errorf("Expect all nodes representing replication controllers are removed from the Propagator's graph")
 	}
+	metric := &dto.Metric{}
+	garbagecollector.EventProcessingLatency.Write(metric)
+	count := float64(metric.Summary.GetSampleCount())
+	sum := metric.Summary.GetSampleSum()
+	t.Logf("Average time spent in GC's eventQueue is %.1f microseconds", sum/count)
+	garbagecollector.DirtyProcessingLatency.Write(metric)
+	count = float64(metric.Summary.GetSampleCount())
+	sum = metric.Summary.GetSampleSum()
+	t.Logf("Average time spent in GC's dirtyQueue is %.1f microseconds", sum/count)
+	garbagecollector.OrphanProcessingLatency.Write(metric)
+	count = float64(metric.Summary.GetSampleCount())
+	sum = metric.Summary.GetSampleSum()
+	t.Logf("Average time spent in GC's orphanQueue is %.1f microseconds", sum/count)
 }
 
 func TestOrphaning(t *testing.T) {