diff --git a/test/e2e_federation/BUILD b/test/e2e_federation/BUILD
index 94545987d60..a20a6a8f5b7 100644
--- a/test/e2e_federation/BUILD
+++ b/test/e2e_federation/BUILD
@@ -49,9 +49,7 @@ go_library(
         "//vendor:k8s.io/apimachinery/pkg/util/rand",
         "//vendor:k8s.io/apimachinery/pkg/util/uuid",
         "//vendor:k8s.io/apimachinery/pkg/util/wait",
-        "//vendor:k8s.io/client-go/rest",
         "//vendor:k8s.io/client-go/tools/clientcmd",
-        "//vendor:k8s.io/client-go/tools/clientcmd/api",
     ],
 )
 
diff --git a/test/e2e_federation/apiserver.go b/test/e2e_federation/apiserver.go
index 90c6412f7a1..f6fbcbf1359 100644
--- a/test/e2e_federation/apiserver.go
+++ b/test/e2e_federation/apiserver.go
@@ -18,14 +18,10 @@ package e2e_federation
 
 import (
 	"fmt"
-	"time"
 
 	. "github.com/onsi/ginkgo"
 	. "github.com/onsi/gomega"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/apimachinery/pkg/util/wait"
-	federationapi "k8s.io/kubernetes/federation/apis/federation/v1beta1"
-	"k8s.io/kubernetes/federation/client/clientset_generated/federation_clientset"
 	"k8s.io/kubernetes/pkg/api/v1"
 	"k8s.io/kubernetes/test/e2e/framework"
 	fedframework "k8s.io/kubernetes/test/e2e_federation/framework"
@@ -61,7 +57,7 @@ var _ = framework.KubeDescribe("Federation apiserver [Feature:Federation]", func
 
 			framework.Logf("Checking that %d clusters are Ready", len(contexts))
 			for _, context := range contexts {
-				clusterIsReadyOrFail(f, &context)
+				fedframework.ClusterIsReadyOrFail(f, &context)
 			}
 			framework.Logf("%d clusters are Ready", len(contexts))
 
@@ -122,19 +118,3 @@ func newService(name, namespace string) *v1.Service {
 		},
 	}
 }
-
-// Verify that the cluster is marked ready.
-func isReady(clusterName string, clientset *federation_clientset.Clientset) error {
-	return wait.PollImmediate(time.Second, 5*time.Minute, func() (bool, error) {
-		c, err := clientset.Federation().Clusters().Get(clusterName, metav1.GetOptions{})
-		if err != nil {
-			return false, err
-		}
-		for _, condition := range c.Status.Conditions {
-			if condition.Type == federationapi.ClusterReady && condition.Status == v1.ConditionTrue {
-				return true, nil
-			}
-		}
-		return false, nil
-	})
-}
diff --git a/test/e2e_federation/daemonset.go b/test/e2e_federation/daemonset.go
index 9fe203504ef..eeaf2a2dad9 100644
--- a/test/e2e_federation/daemonset.go
+++ b/test/e2e_federation/daemonset.go
@@ -43,7 +43,7 @@ const (
 
 // Create/delete daemonset api objects
 var _ = framework.KubeDescribe("Federation daemonsets [Feature:Federation]", func() {
-	var clusters map[string]*cluster // All clusters, keyed by cluster name
+	var clusters fedframework.ClusterSlice
 
 	f := fedframework.NewDefaultFederatedFramework("federated-daemonset")
 
@@ -51,7 +51,7 @@ var _ = framework.KubeDescribe("Federation daemonsets [Feature:Federation]", fun
 
 		BeforeEach(func() {
 			fedframework.SkipUnlessFederated(f.ClientSet)
-			clusters, _ = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		AfterEach(func() {
@@ -114,7 +114,7 @@ func deleteAllDaemonSetsOrFail(clientset *fedclientset.Clientset, nsName string)
 // verifyCascadingDeletionForDS verifies that daemonsets are deleted from
 // underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyCascadingDeletionForDS(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName string) {
+func verifyCascadingDeletionForDS(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName string) {
 	daemonset := createDaemonSetOrFail(clientset, nsName)
 	daemonsetName := daemonset.Name
 	// Check subclusters if the daemonset was created there.
@@ -140,8 +140,9 @@ func verifyCascadingDeletionForDS(clientset *fedclientset.Clientset, clusters ma
 	errMessages := []string{}
 	// daemon set should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Extensions().DaemonSets(nsName).Get(daemonsetName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Extensions().DaemonSets(nsName).Get(daemonsetName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for daemonset %s in cluster %s, expected daemonset to exist", daemonsetName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
@@ -233,7 +234,7 @@ func updateDaemonSetOrFail(clientset *fedclientset.Clientset, namespace string)
 	return newDaemonSet
 }
 
-func waitForDaemonSetShardsOrFail(namespace string, daemonset *v1beta1.DaemonSet, clusters map[string]*cluster) {
+func waitForDaemonSetShardsOrFail(namespace string, daemonset *v1beta1.DaemonSet, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for daemonset %q in %d clusters", daemonset.Name, len(clusters))
 	for _, c := range clusters {
 		waitForDaemonSetOrFail(c.Clientset, namespace, daemonset, true, FederatedDaemonSetTimeout)
@@ -263,7 +264,7 @@ func waitForDaemonSetOrFail(clientset *kubeclientset.Clientset, namespace string
 	}
 }
 
-func waitForDaemonSetShardsUpdatedOrFail(namespace string, daemonset *v1beta1.DaemonSet, clusters map[string]*cluster) {
+func waitForDaemonSetShardsUpdatedOrFail(namespace string, daemonset *v1beta1.DaemonSet, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for daemonset %q in %d clusters", daemonset.Name, len(clusters))
 	for _, c := range clusters {
 		waitForDaemonSetUpdateOrFail(c.Clientset, namespace, daemonset, FederatedDaemonSetTimeout)
diff --git a/test/e2e_federation/deployment.go b/test/e2e_federation/deployment.go
index 2a7c663cd85..e873dcd8986 100644
--- a/test/e2e_federation/deployment.go
+++ b/test/e2e_federation/deployment.go
@@ -69,11 +69,11 @@ var _ = framework.KubeDescribe("Federation deployments [Feature:Federation]", fu
 	// e2e cases for federated deployment controller
 	Describe("Federated Deployment", func() {
 		var (
-			clusters map[string]*cluster
+			clusters fedframework.ClusterSlice
 		)
 		BeforeEach(func() {
 			fedframework.SkipUnlessFederated(f.ClientSet)
-			clusters, _ = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		AfterEach(func() {
@@ -140,7 +140,7 @@ func deleteAllDeploymentsOrFail(clientset *fedclientset.Clientset, nsName string
 // verifyCascadingDeletionForDeployment verifies that deployments are deleted
 // from underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyCascadingDeletionForDeployment(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName string) {
+func verifyCascadingDeletionForDeployment(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName string) {
 	deployment := createDeploymentOrFail(clientset, nsName)
 	deploymentName := deployment.Name
 	// Check subclusters if the deployment was created there.
@@ -166,8 +166,9 @@ func verifyCascadingDeletionForDeployment(clientset *fedclientset.Clientset, clu
 	errMessages := []string{}
 	// deployment should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Extensions().Deployments(nsName).Get(deploymentName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Extensions().Deployments(nsName).Get(deploymentName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for deployment %s in cluster %s, expected deployment to exist", deploymentName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
@@ -179,13 +180,13 @@ func verifyCascadingDeletionForDeployment(clientset *fedclientset.Clientset, clu
 	}
 }
 
-func waitForDeploymentOrFail(c *fedclientset.Clientset, namespace string, deploymentName string, clusters map[string]*cluster) {
+func waitForDeploymentOrFail(c *fedclientset.Clientset, namespace string, deploymentName string, clusters fedframework.ClusterSlice) {
 	err := waitForDeployment(c, namespace, deploymentName, clusters)
 	framework.ExpectNoError(err, "Failed to verify deployment %q/%q, err: %v", namespace, deploymentName, err)
 }
 
-func waitForDeployment(c *fedclientset.Clientset, namespace string, deploymentName string, clusters map[string]*cluster) error {
-	err := wait.Poll(10*time.Second, federatedDefaultTestTimeout, func() (bool, error) {
+func waitForDeployment(c *fedclientset.Clientset, namespace string, deploymentName string, clusters fedframework.ClusterSlice) error {
+	err := wait.Poll(10*time.Second, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 		fdep, err := c.Deployments(namespace).Get(deploymentName, metav1.GetOptions{})
 		if err != nil {
 			return false, err
@@ -194,12 +195,12 @@ func waitForDeployment(c *fedclientset.Clientset, namespace string, deploymentNa
 		for _, cluster := range clusters {
 			dep, err := cluster.Extensions().Deployments(namespace).Get(deploymentName, metav1.GetOptions{})
 			if err != nil && !errors.IsNotFound(err) {
-				By(fmt.Sprintf("Failed getting deployment: %q/%q/%q, err: %v", cluster.name, namespace, deploymentName, err))
+				By(fmt.Sprintf("Failed getting deployment: %q/%q/%q, err: %v", cluster.Name, namespace, deploymentName, err))
 				return false, err
 			}
 			if err == nil {
 				if !verifyDeployment(fdep, dep) {
-					By(fmt.Sprintf("Deployment meta or spec not match for cluster %q:\n    federation: %v\n    cluster: %v", cluster.name, fdep, dep))
+					By(fmt.Sprintf("Deployment meta or spec not match for cluster %q:\n    federation: %v\n    cluster: %v", cluster.Name, fdep, dep))
 					return false, nil
 				}
 				specReplicas += *dep.Spec.Replicas
@@ -259,7 +260,7 @@ func deleteDeploymentOrFail(clientset *fedclientset.Clientset, nsName string, de
 	}
 
 	// Wait for the deployment to be deleted.
-	err = wait.Poll(10*time.Second, federatedDefaultTestTimeout, func() (bool, error) {
+	err = wait.Poll(10*time.Second, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 		_, err := clientset.Extensions().Deployments(nsName).Get(deploymentName, metav1.GetOptions{})
 		if err != nil && errors.IsNotFound(err) {
 			return true, nil
diff --git a/test/e2e_federation/framework/BUILD b/test/e2e_federation/framework/BUILD
index 3d98d57591d..74e050070ae 100644
--- a/test/e2e_federation/framework/BUILD
+++ b/test/e2e_federation/framework/BUILD
@@ -10,6 +10,7 @@ load(
 go_library(
     name = "go_default_library",
     srcs = [
+        "cluster.go",
         "framework.go",
         "util.go",
     ],
@@ -31,6 +32,7 @@ go_library(
         "//vendor:k8s.io/apimachinery/pkg/util/wait",
         "//vendor:k8s.io/client-go/rest",
         "//vendor:k8s.io/client-go/tools/clientcmd",
+        "//vendor:k8s.io/client-go/tools/clientcmd/api",
     ],
 )
 
diff --git a/test/e2e_federation/framework/cluster.go b/test/e2e_federation/framework/cluster.go
new file mode 100644
index 00000000000..d5dba98dd3b
--- /dev/null
+++ b/test/e2e_federation/framework/cluster.go
@@ -0,0 +1,164 @@
+/*
+Copyright 2017 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 framework
+
+import (
+	"fmt"
+	"time"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/wait"
+	restclient "k8s.io/client-go/rest"
+	"k8s.io/client-go/tools/clientcmd"
+	clientcmdapi "k8s.io/client-go/tools/clientcmd/api"
+	federationapi "k8s.io/kubernetes/federation/apis/federation/v1beta1"
+	fedclientset "k8s.io/kubernetes/federation/client/clientset_generated/federation_clientset"
+	"k8s.io/kubernetes/pkg/api/v1"
+	kubeclientset "k8s.io/kubernetes/pkg/client/clientset_generated/clientset"
+	"k8s.io/kubernetes/test/e2e/framework"
+
+	. "github.com/onsi/ginkgo"
+	. "github.com/onsi/gomega"
+)
+
+const (
+	KubeAPIQPS   float32 = 20.0
+	KubeAPIBurst         = 30
+
+	userAgentName = "federation-e2e"
+
+	federatedNamespaceTimeout    = 5 * time.Minute
+	federatedClustersWaitTimeout = 1 * time.Minute
+)
+
+// ClusterSlice is a slice of clusters
+type ClusterSlice []*Cluster
+
+// Cluster keeps track of the name and client of a cluster in the federation
+type Cluster struct {
+	Name string
+	*kubeclientset.Clientset
+}
+
+func getRegisteredClusters(f *Framework) ClusterSlice {
+	contexts := f.GetUnderlyingFederatedContexts()
+
+	By("Obtaining a list of all the clusters")
+	clusterList := waitForAllRegisteredClusters(f, len(contexts))
+
+	framework.Logf("Checking that %d clusters are Ready", len(contexts))
+	for _, context := range contexts {
+		ClusterIsReadyOrFail(f, &context)
+	}
+	framework.Logf("%d clusters are Ready", len(contexts))
+
+	clusters := ClusterSlice{}
+	for i, c := range clusterList.Items {
+		framework.Logf("Creating a clientset for the cluster %s", c.Name)
+		Expect(framework.TestContext.KubeConfig).ToNot(Equal(""), "KubeConfig must be specified to load clusters' client config")
+		clusters = append(clusters, &Cluster{
+			Name:      c.Name,
+			Clientset: createClientsetForCluster(c, i, userAgentName),
+		})
+	}
+	waitForNamespaceInFederatedClusters(clusters, f.FederationNamespace.Name, federatedNamespaceTimeout)
+	return clusters
+}
+
+// waitForAllRegisteredClusters waits for all clusters defined in e2e context to be created
+// return ClusterList until the listed cluster items equals clusterCount
+func waitForAllRegisteredClusters(f *Framework, clusterCount int) *federationapi.ClusterList {
+	var clusterList *federationapi.ClusterList
+	if err := wait.PollImmediate(framework.Poll, federatedClustersWaitTimeout, func() (bool, error) {
+		var err error
+		clusterList, err = f.FederationClientset.Federation().Clusters().List(metav1.ListOptions{})
+		if err != nil {
+			return false, err
+		}
+		framework.Logf("%d clusters registered, waiting for %d", len(clusterList.Items), clusterCount)
+		if len(clusterList.Items) == clusterCount {
+			return true, nil
+		}
+		return false, nil
+	}); err != nil {
+		framework.Failf("Failed to list registered clusters: %+v", err)
+	}
+	return clusterList
+}
+
+func createClientsetForCluster(c federationapi.Cluster, i int, userAgentName string) *kubeclientset.Clientset {
+	kubecfg, err := clientcmd.LoadFromFile(framework.TestContext.KubeConfig)
+	framework.ExpectNoError(err, "error loading KubeConfig: %v", err)
+
+	cfgOverride := &clientcmd.ConfigOverrides{
+		ClusterInfo: clientcmdapi.Cluster{
+			Server: c.Spec.ServerAddressByClientCIDRs[0].ServerAddress,
+		},
+	}
+	ccfg := clientcmd.NewNonInteractiveClientConfig(*kubecfg, c.Name, cfgOverride, clientcmd.NewDefaultClientConfigLoadingRules())
+	cfg, err := ccfg.ClientConfig()
+	framework.ExpectNoError(err, "Error creating client config in cluster #%d (%q)", i, c.Name)
+
+	cfg.QPS = KubeAPIQPS
+	cfg.Burst = KubeAPIBurst
+	return kubeclientset.NewForConfigOrDie(restclient.AddUserAgent(cfg, userAgentName))
+}
+
+// waitForNamespaceInFederatedClusters waits for the federated namespace to be created in federated clusters
+func waitForNamespaceInFederatedClusters(clusters ClusterSlice, nsName string, timeout time.Duration) {
+	for _, c := range clusters {
+		name := c.Name
+		err := wait.PollImmediate(framework.Poll, timeout, func() (bool, error) {
+			_, err := c.Clientset.Core().Namespaces().Get(nsName, metav1.GetOptions{})
+			if err != nil {
+				By(fmt.Sprintf("Waiting for namespace %q to be created in cluster %q, err: %v", nsName, name, err))
+				return false, nil
+			}
+			By(fmt.Sprintf("Namespace %q exists in cluster %q", nsName, name))
+			return true, nil
+		})
+		framework.ExpectNoError(err, "Failed to verify federated namespace %q creation in cluster %q", nsName, name)
+	}
+}
+
+// ClusterIsReadyOrFail checks whether the federated cluster of the provided context is ready
+func ClusterIsReadyOrFail(f *Framework, context *E2EContext) {
+	c, err := f.FederationClientset.Federation().Clusters().Get(context.Name, metav1.GetOptions{})
+	framework.ExpectNoError(err, fmt.Sprintf("get cluster: %+v", err))
+	if c.ObjectMeta.Name != context.Name {
+		framework.Failf("cluster name does not match input context: actual=%+v, expected=%+v", c, context)
+	}
+	err = isReady(context.Name, f.FederationClientset)
+	framework.ExpectNoError(err, fmt.Sprintf("unexpected error in verifying if cluster %s is ready: %+v", context.Name, err))
+	framework.Logf("Cluster %s is Ready", context.Name)
+}
+
+// Verify that the cluster is marked ready.
+func isReady(clusterName string, clientset *fedclientset.Clientset) error {
+	return wait.PollImmediate(time.Second, 5*time.Minute, func() (bool, error) {
+		c, err := clientset.Federation().Clusters().Get(clusterName, metav1.GetOptions{})
+		if err != nil {
+			return false, err
+		}
+		for _, condition := range c.Status.Conditions {
+			if condition.Type == federationapi.ClusterReady && condition.Status == v1.ConditionTrue {
+				return true, nil
+			}
+		}
+		return false, nil
+	})
+}
diff --git a/test/e2e_federation/framework/framework.go b/test/e2e_federation/framework/framework.go
index 2f38e06e881..38c5fb20560 100644
--- a/test/e2e_federation/framework/framework.go
+++ b/test/e2e_federation/framework/framework.go
@@ -231,3 +231,7 @@ func (f *Framework) GetUnderlyingFederatedContexts() []E2EContext {
 
 	return e2eContexts
 }
+
+func (f *Framework) GetRegisteredClusters() ClusterSlice {
+	return getRegisteredClusters(f)
+}
diff --git a/test/e2e_federation/framework/util.go b/test/e2e_federation/framework/util.go
index 2daea426499..0d9b1673867 100644
--- a/test/e2e_federation/framework/util.go
+++ b/test/e2e_federation/framework/util.go
@@ -37,6 +37,8 @@ import (
 	"k8s.io/kubernetes/test/e2e/framework"
 )
 
+const FederatedDefaultTestTimeout = 5 * time.Minute
+
 // Detects whether the federation namespace exists in the underlying cluster
 func SkipUnlessFederated(c clientset.Interface) {
 	federationNS := framework.FederationSystemNamespace()
diff --git a/test/e2e_federation/ingress.go b/test/e2e_federation/ingress.go
index b73787372a6..701dd2583f5 100644
--- a/test/e2e_federation/ingress.go
+++ b/test/e2e_federation/ingress.go
@@ -142,11 +142,11 @@ var _ = framework.KubeDescribe("Federated ingresses [Feature:Federation]", func(
 	// e2e cases for federation ingress controller
 	var _ = Describe("Federated Ingresses", func() {
 		var (
-			clusters                               map[string]*cluster // All clusters, keyed by cluster name
-			primaryClusterName, federationName, ns string
-			jig                                    *federationTestJig
-			service                                *v1.Service
-			secret                                 *v1.Secret
+			clusters           fedframework.ClusterSlice
+			federationName, ns string
+			jig                *federationTestJig
+			service            *v1.Service
+			secret             *v1.Secret
 		)
 
 		// register clusters in federation apiserver
@@ -157,7 +157,7 @@ var _ = framework.KubeDescribe("Federated ingresses [Feature:Federation]", func(
 				federationName = DefaultFederationName
 			}
 			jig = newFederationTestJig(f.FederationClientset)
-			clusters, primaryClusterName = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 			ns = f.FederationNamespace.Name
 			// create backend service
 			service = createServiceOrFail(f.FederationClientset, ns, FederatedIngressServiceName)
@@ -221,10 +221,12 @@ var _ = framework.KubeDescribe("Federated ingresses [Feature:Federation]", func(
 
 		var _ = Describe("Ingress connectivity and DNS", func() {
 
+			var backendPods BackendPodMap
+
 			BeforeEach(func() {
 				fedframework.SkipUnlessFederated(f.ClientSet)
 				// create backend pod
-				createBackendPodsOrFail(clusters, ns, FederatedIngressServicePodName)
+				backendPods = createBackendPodsOrFail(clusters, ns, FederatedIngressServicePodName)
 				// create ingress object
 				jig.ing = createIngressOrFail(f.FederationClientset, ns, service.Name, FederatedIngressTLSSecretName)
 				// wait for ingress objects sync
@@ -233,7 +235,9 @@ var _ = framework.KubeDescribe("Federated ingresses [Feature:Federation]", func(
 			})
 
 			AfterEach(func() {
-				deleteBackendPodsOrFail(clusters, ns)
+				deleteBackendPodsOrFail(clusters, backendPods)
+				backendPods = nil
+
 				if jig.ing != nil {
 					By(fmt.Sprintf("Deleting ingress %v on all clusters", jig.ing.Name))
 					orphanDependents := false
@@ -285,7 +289,7 @@ func equivalentIngress(federatedIngress, clusterIngress v1beta1.Ingress) bool {
 // verifyCascadingDeletionForIngress verifies that ingresses are deleted from
 // underlying clusters when orphan dependents is false and they are not deleted
 // when orphan dependents is true.
-func verifyCascadingDeletionForIngress(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName string) {
+func verifyCascadingDeletionForIngress(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName string) {
 	ingress := createIngressOrFail(clientset, nsName, FederatedIngressServiceName, FederatedIngressTLSSecretName)
 	ingressName := ingress.Name
 	// Check subclusters if the ingress was created there.
@@ -299,8 +303,9 @@ func verifyCascadingDeletionForIngress(clientset *fedclientset.Clientset, cluste
 	errMessages := []string{}
 	// ingress should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Extensions().Ingresses(nsName).Get(ingressName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Extensions().Ingresses(nsName).Get(ingressName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for ingress %s in cluster %s, expected ingress to exist", ingressName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
@@ -338,7 +343,7 @@ func waitForIngressOrFail(clientset *kubeclientset.Clientset, namespace string,
 }
 
 // waitForIngressShardsOrFail waits for the ingress to appear in all clusters
-func waitForIngressShardsOrFail(namespace string, ingress *v1beta1.Ingress, clusters map[string]*cluster) {
+func waitForIngressShardsOrFail(namespace string, ingress *v1beta1.Ingress, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for ingress %q in %d clusters", ingress.Name, len(clusters))
 	for _, c := range clusters {
 		waitForIngressOrFail(c.Clientset, namespace, ingress, true, FederatedIngressTimeout)
@@ -346,7 +351,7 @@ func waitForIngressShardsOrFail(namespace string, ingress *v1beta1.Ingress, clus
 }
 
 // waitForIngressShardsUpdatedOrFail waits for the ingress to be updated in all clusters
-func waitForIngressShardsUpdatedOrFail(namespace string, ingress *v1beta1.Ingress, clusters map[string]*cluster) {
+func waitForIngressShardsUpdatedOrFail(namespace string, ingress *v1beta1.Ingress, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for ingress %q in %d clusters", ingress.Name, len(clusters))
 	for _, c := range clusters {
 		waitForIngressUpdateOrFail(c.Clientset, namespace, ingress, FederatedIngressTimeout)
@@ -374,7 +379,7 @@ func waitForIngressUpdateOrFail(clientset *kubeclientset.Clientset, namespace st
 }
 
 // waitForIngressShardsGoneOrFail waits for the ingress to disappear in all clusters
-func waitForIngressShardsGoneOrFail(namespace string, ingress *v1beta1.Ingress, clusters map[string]*cluster) {
+func waitForIngressShardsGoneOrFail(namespace string, ingress *v1beta1.Ingress, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for ingress %q in %d clusters", ingress.Name, len(clusters))
 	for _, c := range clusters {
 		waitForIngressOrFail(c.Clientset, namespace, ingress, false, FederatedIngressTimeout)
diff --git a/test/e2e_federation/namespace.go b/test/e2e_federation/namespace.go
index a5e0c220507..69569d78820 100644
--- a/test/e2e_federation/namespace.go
+++ b/test/e2e_federation/namespace.go
@@ -44,13 +44,13 @@ var _ = framework.KubeDescribe("Federation namespace [Feature:Federation]", func
 	f := fedframework.NewDefaultFederatedFramework("federation-namespace")
 
 	Describe("Namespace objects", func() {
-		var clusters map[string]*cluster // All clusters, keyed by cluster name
+		var clusters fedframework.ClusterSlice
 
 		var nsName string
 
 		BeforeEach(func() {
 			fedframework.SkipUnlessFederated(f.ClientSet)
-			clusters, _ = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		AfterEach(func() {
@@ -188,7 +188,7 @@ var _ = framework.KubeDescribe("Federation namespace [Feature:Federation]", func
 // verifyNsCascadingDeletion verifies that namespaces are deleted from
 // underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyNsCascadingDeletion(nsClient clientset.NamespaceInterface, clusters map[string]*cluster, orphanDependents *bool) string {
+func verifyNsCascadingDeletion(nsClient clientset.NamespaceInterface, clusters fedframework.ClusterSlice, orphanDependents *bool) string {
 	nsName := createNamespace(nsClient)
 	// Check subclusters if the namespace was created there.
 	By(fmt.Sprintf("Waiting for namespace %s to be created in all underlying clusters", nsName))
@@ -213,8 +213,9 @@ func verifyNsCascadingDeletion(nsClient clientset.NamespaceInterface, clusters m
 	errMessages := []string{}
 	// namespace should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Core().Namespaces().Get(nsName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Core().Namespaces().Get(nsName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for namespace %s in cluster %s, expected namespace to exist", nsName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
diff --git a/test/e2e_federation/replicaset.go b/test/e2e_federation/replicaset.go
index 3d108f6d120..f542449d9c2 100644
--- a/test/e2e_federation/replicaset.go
+++ b/test/e2e_federation/replicaset.go
@@ -74,12 +74,12 @@ var _ = framework.KubeDescribe("Federated ReplicaSet [Feature:Federation]", func
 	// e2e cases for federated replicaset controller
 	Describe("Features", func() {
 		var (
-			clusters map[string]*cluster
+			clusters fedframework.ClusterSlice
 		)
 
 		BeforeEach(func() {
 			fedframework.SkipUnlessFederated(f.ClientSet)
-			clusters, _ = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		// e2e cases for federated replicaset controller
@@ -200,7 +200,7 @@ var _ = framework.KubeDescribe("Federated ReplicaSet [Feature:Federation]", func
 	})
 })
 
-func createAndWaitForReplicasetOrFail(clientset *fedclientset.Clientset, nsName string, clusters map[string]*cluster) *v1beta1.ReplicaSet {
+func createAndWaitForReplicasetOrFail(clientset *fedclientset.Clientset, nsName string, clusters fedframework.ClusterSlice) *v1beta1.ReplicaSet {
 	rs := createReplicaSetOrFail(clientset, newReplicaSet(nsName, FederationReplicaSetPrefix, 5, nil))
 	// Check subclusters if the replicaSet was created there.
 	By(fmt.Sprintf("Waiting for replica sets %s to be created in all underlying clusters", rs.Name))
@@ -220,7 +220,7 @@ func createAndWaitForReplicasetOrFail(clientset *fedclientset.Clientset, nsName
 	return rs
 }
 
-func createAndUpdateFedRSWithPref(clientset *fedclientset.Clientset, nsName string, clusters map[string]*cluster, pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) *v1beta1.ReplicaSet {
+func createAndUpdateFedRSWithPref(clientset *fedclientset.Clientset, nsName string, clusters fedframework.ClusterSlice, pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) *v1beta1.ReplicaSet {
 	framework.Logf("Replicas: %d, Preference: %#v", replicas, pref)
 	rs := newReplicaSet(nsName, FederationReplicaSetPrefix, replicas, pref)
 	rs = createReplicaSetOrFail(clientset, rs)
@@ -254,14 +254,15 @@ func deleteAllReplicaSetsOrFail(clientset *fedclientset.Clientset, nsName string
 // verifyCascadingDeletionForReplicaSet verifies that replicaSets are deleted
 // from underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyCascadingDeletionForReplicaSet(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName, rsName string) {
+func verifyCascadingDeletionForReplicaSet(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName, rsName string) {
 	By(fmt.Sprintf("Deleting replica set %s", rsName))
 	deleteReplicaSetOrFail(clientset, nsName, rsName, orphanDependents)
 
 	By(fmt.Sprintf("Verifying replica sets %s in underlying clusters", rsName))
 	errMessages := []string{}
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Extensions().ReplicaSets(nsName).Get(rsName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Extensions().ReplicaSets(nsName).Get(rsName, metav1.GetOptions{})
 		if (orphanDependents == nil || *orphanDependents == true) && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for replica set %s in cluster %s, expected replica set to exist", rsName, clusterName))
 		} else if (orphanDependents != nil && *orphanDependents == false) && (err == nil || !errors.IsNotFound(err)) {
@@ -273,7 +274,7 @@ func verifyCascadingDeletionForReplicaSet(clientset *fedclientset.Clientset, clu
 	}
 }
 
-func generateFedRSPrefsWithWeight(clusters map[string]*cluster) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
+func generateFedRSPrefsWithWeight(clusters fedframework.ClusterSlice) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
 	By("Generating replicaset preferences with weights")
 	clusterNames := extractClusterNames(clusters)
 	pref = &federation.FederatedReplicaSetPreferences{
@@ -294,7 +295,7 @@ func generateFedRSPrefsWithWeight(clusters map[string]*cluster) (pref *federatio
 	return
 }
 
-func generateFedRSPrefsWithMin(clusters map[string]*cluster) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
+func generateFedRSPrefsWithMin(clusters fedframework.ClusterSlice) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
 	By("Generating replicaset preferences with min replicas")
 	clusterNames := extractClusterNames(clusters)
 	pref = &federation.FederatedReplicaSetPreferences{
@@ -321,7 +322,7 @@ func generateFedRSPrefsWithMin(clusters map[string]*cluster) (pref *federation.F
 	return
 }
 
-func generateFedRSPrefsWithMax(clusters map[string]*cluster) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
+func generateFedRSPrefsWithMax(clusters fedframework.ClusterSlice) (pref *federation.FederatedReplicaSetPreferences, replicas int32, expect map[string]int32) {
 	By("Generating replicaset preferences with max replicas")
 	clusterNames := extractClusterNames(clusters)
 	pref = &federation.FederatedReplicaSetPreferences{
@@ -354,7 +355,7 @@ func updateFedRSPrefsRebalance(pref *federation.FederatedReplicaSetPreferences,
 	return pref
 }
 
-func generateFedRSPrefsForRebalancing(clusters map[string]*cluster) (pref1, pref2 *federation.FederatedReplicaSetPreferences, replicas int32, expect1, expect2 map[string]int32) {
+func generateFedRSPrefsForRebalancing(clusters fedframework.ClusterSlice) (pref1, pref2 *federation.FederatedReplicaSetPreferences, replicas int32, expect1, expect2 map[string]int32) {
 	By("Generating replicaset for rebalancing")
 	clusterNames := extractClusterNames(clusters)
 	replicas = 3
@@ -382,14 +383,14 @@ func generateFedRSPrefsForRebalancing(clusters map[string]*cluster) (pref1, pref
 	return
 }
 
-func waitForReplicaSetOrFail(c *fedclientset.Clientset, namespace string, replicaSetName string, clusters map[string]*cluster, expect map[string]int32) {
+func waitForReplicaSetOrFail(c *fedclientset.Clientset, namespace string, replicaSetName string, clusters fedframework.ClusterSlice, expect map[string]int32) {
 	err := waitForReplicaSet(c, namespace, replicaSetName, clusters, expect)
 	framework.ExpectNoError(err, "Failed to verify replica set \"%s/%s\", err: %v", namespace, replicaSetName, err)
 }
 
-func waitForReplicaSet(c *fedclientset.Clientset, namespace string, replicaSetName string, clusters map[string]*cluster, expect map[string]int32) error {
+func waitForReplicaSet(c *fedclientset.Clientset, namespace string, replicaSetName string, clusters fedframework.ClusterSlice, expect map[string]int32) error {
 	framework.Logf("waitForReplicaSet: %s/%s; clusters: %v; expect: %v", namespace, replicaSetName, clusters, expect)
-	err := wait.Poll(10*time.Second, federatedDefaultTestTimeout, func() (bool, error) {
+	err := wait.Poll(10*time.Second, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 		frs, err := c.ReplicaSets(namespace).Get(replicaSetName, metav1.GetOptions{})
 		if err != nil {
 			return false, err
@@ -398,21 +399,21 @@ func waitForReplicaSet(c *fedclientset.Clientset, namespace string, replicaSetNa
 		for _, cluster := range clusters {
 			rs, err := cluster.ReplicaSets(namespace).Get(replicaSetName, metav1.GetOptions{})
 			if err != nil && !errors.IsNotFound(err) {
-				framework.Logf("Failed getting replicaset: \"%s/%s/%s\", err: %v", cluster.name, namespace, replicaSetName, err)
+				framework.Logf("Failed getting replicaset: \"%s/%s/%s\", err: %v", cluster.Name, namespace, replicaSetName, err)
 				return false, err
 			}
 			if errors.IsNotFound(err) {
-				if expect != nil && expect[cluster.name] > 0 {
-					framework.Logf("Replicaset \"%s/%s/%s\" with replica count %d does not exist", cluster.name, namespace, replicaSetName, expect[cluster.name])
+				if expect != nil && expect[cluster.Name] > 0 {
+					framework.Logf("Replicaset \"%s/%s/%s\" with replica count %d does not exist", cluster.Name, namespace, replicaSetName, expect[cluster.Name])
 					return false, nil
 				}
 			} else {
 				if !equivalentReplicaSet(frs, rs) {
-					framework.Logf("Replicaset meta or spec does not match for cluster %q:\n    federation: %v\n    cluster: %v", cluster.name, frs, rs)
+					framework.Logf("Replicaset meta or spec does not match for cluster %q:\n    federation: %v\n    cluster: %v", cluster.Name, frs, rs)
 					return false, nil
 				}
-				if expect != nil && *rs.Spec.Replicas < expect[cluster.name] {
-					framework.Logf("Replicas do not match for \"%s/%s/%s\": expected: >= %v, actual: %v", cluster.name, namespace, replicaSetName, expect[cluster.name], *rs.Spec.Replicas)
+				if expect != nil && *rs.Spec.Replicas < expect[cluster.Name] {
+					framework.Logf("Replicas do not match for \"%s/%s/%s\": expected: >= %v, actual: %v", cluster.Name, namespace, replicaSetName, expect[cluster.Name], *rs.Spec.Replicas)
 					return false, nil
 				}
 				specReplicas += *rs.Spec.Replicas
@@ -530,10 +531,10 @@ func newReplicaSetWithName(namespace string, name string, replicas int32, pref *
 	return rs
 }
 
-func extractClusterNames(clusters map[string]*cluster) []string {
+func extractClusterNames(clusters fedframework.ClusterSlice) []string {
 	clusterNames := make([]string, 0, len(clusters))
-	for clusterName := range clusters {
-		clusterNames = append(clusterNames, clusterName)
+	for _, cluster := range clusters {
+		clusterNames = append(clusterNames, cluster.Name)
 	}
 	return clusterNames
 }
diff --git a/test/e2e_federation/secret.go b/test/e2e_federation/secret.go
index cd929519770..40e8c414835 100644
--- a/test/e2e_federation/secret.go
+++ b/test/e2e_federation/secret.go
@@ -42,7 +42,7 @@ const (
 
 // Create/delete secret api objects
 var _ = framework.KubeDescribe("Federation secrets [Feature:Federation]", func() {
-	var clusters map[string]*cluster // All clusters, keyed by cluster name
+	var clusters fedframework.ClusterSlice
 
 	f := fedframework.NewDefaultFederatedFramework("federated-secret")
 
@@ -50,7 +50,7 @@ var _ = framework.KubeDescribe("Federation secrets [Feature:Federation]", func()
 
 		BeforeEach(func() {
 			fedframework.SkipUnlessFederated(f.ClientSet)
-			clusters, _ = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		AfterEach(func() {
@@ -108,7 +108,7 @@ func deleteAllSecretsOrFail(clientset *fedclientset.Clientset, nsName string) {
 // verifyCascadingDeletionForSecret verifies that secrets are deleted from
 // underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyCascadingDeletionForSecret(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName string) {
+func verifyCascadingDeletionForSecret(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName string) {
 	secret := createSecretOrFail(clientset, nsName)
 	secretName := secret.Name
 	// Check subclusters if the secret was created there.
@@ -134,8 +134,9 @@ func verifyCascadingDeletionForSecret(clientset *fedclientset.Clientset, cluster
 	errMessages := []string{}
 	// secret should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Core().Secrets(nsName).Get(secretName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Core().Secrets(nsName).Get(secretName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for secret %s in cluster %s, expected secret to exist", secretName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
@@ -213,7 +214,7 @@ func updateSecretOrFail(clientset *fedclientset.Clientset, nsName string, secret
 	return newSecret
 }
 
-func waitForSecretShardsOrFail(nsName string, secret *v1.Secret, clusters map[string]*cluster) {
+func waitForSecretShardsOrFail(nsName string, secret *v1.Secret, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for secret %q in %d clusters", secret.Name, len(clusters))
 	for _, c := range clusters {
 		waitForSecretOrFail(c.Clientset, nsName, secret, true, FederatedSecretTimeout)
@@ -243,7 +244,7 @@ func waitForSecretOrFail(clientset *kubeclientset.Clientset, nsName string, secr
 	}
 }
 
-func waitForSecretShardsUpdatedOrFail(nsName string, secret *v1.Secret, clusters map[string]*cluster) {
+func waitForSecretShardsUpdatedOrFail(nsName string, secret *v1.Secret, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for secret %q in %d clusters", secret.Name, len(clusters))
 	for _, c := range clusters {
 		waitForSecretUpdateOrFail(c.Clientset, nsName, secret, FederatedSecretTimeout)
diff --git a/test/e2e_federation/service.go b/test/e2e_federation/service.go
index baa67cbf776..479a5fd7878 100644
--- a/test/e2e_federation/service.go
+++ b/test/e2e_federation/service.go
@@ -47,9 +47,8 @@ var FederatedServiceLabels = map[string]string{
 
 var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func() {
 	f := fedframework.NewDefaultFederatedFramework("federated-service")
-	var clusters map[string]*cluster // All clusters, keyed by cluster name
+	var clusters fedframework.ClusterSlice
 	var federationName string
-	var primaryClusterName string // The name of the "primary" cluster
 
 	var _ = Describe("Without Clusters [NoCluster]", func() {
 		BeforeEach(func() {
@@ -84,7 +83,7 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 				federationName = DefaultFederationName
 			}
 
-			clusters, primaryClusterName = getRegisteredClusters(UserAgentName, f)
+			clusters = f.GetRegisteredClusters()
 		})
 
 		Describe("Federated Service", func() {
@@ -146,6 +145,7 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 			var (
 				service      *v1.Service
 				serviceShard *v1.Service
+				backendPods  BackendPodMap
 			)
 
 			BeforeEach(func() {
@@ -153,7 +153,7 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 
 				nsName := f.FederationNamespace.Name
 
-				createBackendPodsOrFail(clusters, nsName, FederatedServicePodName)
+				backendPods = createBackendPodsOrFail(clusters, nsName, FederatedServicePodName)
 
 				service = createServiceOrFail(f.FederationClientset, nsName, FederatedServiceName)
 				obj, err := api.Scheme.DeepCopy(service)
@@ -191,7 +191,8 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 				fedframework.SkipUnlessFederated(f.ClientSet)
 
 				nsName := f.FederationNamespace.Name
-				deleteBackendPodsOrFail(clusters, nsName)
+				deleteBackendPodsOrFail(clusters, backendPods)
+				backendPods = nil
 
 				if service != nil {
 					deleteServiceOrFail(f.FederationClientset, nsName, service.Name, nil)
@@ -244,8 +245,10 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 				BeforeEach(func() {
 					fedframework.SkipUnlessFederated(f.ClientSet)
 
-					// Delete all the backend pods from the shard which is local to the discovery pod.
-					deleteOneBackendPodOrFail(clusters[primaryClusterName])
+					// Delete the backend pod from the shard which is local to the discovery pod.
+					primaryCluster := clusters[0]
+					backendPod := backendPods[primaryCluster.Name]
+					deleteOneBackendPodOrFail(primaryCluster, backendPod)
 
 				})
 
@@ -287,7 +290,7 @@ var _ = framework.KubeDescribe("Federated Services [Feature:Federation]", func()
 // verifyCascadingDeletionForService verifies that services are deleted from
 // underlying clusters when orphan dependents is false and they are not
 // deleted when orphan dependents is true.
-func verifyCascadingDeletionForService(clientset *fedclientset.Clientset, clusters map[string]*cluster, orphanDependents *bool, nsName string) {
+func verifyCascadingDeletionForService(clientset *fedclientset.Clientset, clusters fedframework.ClusterSlice, orphanDependents *bool, nsName string) {
 	service := createServiceOrFail(clientset, nsName, FederatedServiceName)
 	serviceName := service.Name
 	// Check subclusters if the service was created there.
@@ -313,8 +316,9 @@ func verifyCascadingDeletionForService(clientset *fedclientset.Clientset, cluste
 	errMessages := []string{}
 	// service should be present in underlying clusters unless orphanDependents is false.
 	shouldExist := orphanDependents == nil || *orphanDependents == true
-	for clusterName, clusterClientset := range clusters {
-		_, err := clusterClientset.Core().Services(nsName).Get(serviceName, metav1.GetOptions{})
+	for _, cluster := range clusters {
+		clusterName := cluster.Name
+		_, err := cluster.Core().Services(nsName).Get(serviceName, metav1.GetOptions{})
 		if shouldExist && errors.IsNotFound(err) {
 			errMessages = append(errMessages, fmt.Sprintf("unexpected NotFound error for service %s in cluster %s, expected service to exist", serviceName, clusterName))
 		} else if !shouldExist && !errors.IsNotFound(err) {
diff --git a/test/e2e_federation/upgrade.go b/test/e2e_federation/upgrade.go
index ce24bf8b451..6b893067e1a 100644
--- a/test/e2e_federation/upgrade.go
+++ b/test/e2e_federation/upgrade.go
@@ -129,9 +129,9 @@ func federationControlPlaneUpgrade(f *fedframework.Framework) {
 func federatedClustersUpgrade(f *fedframework.Framework) {
 	k8sVersion, err := framework.RealVersion(framework.TestContext.UpgradeTarget)
 	framework.ExpectNoError(err)
-	clusters, _ := getRegisteredClusters(UserAgentName, f)
+	clusters := f.GetRegisteredClusters()
 	for _, cluster := range clusters {
-		framework.ExpectNoError(fedframework.MasterUpgrade(cluster.name, k8sVersion))
+		framework.ExpectNoError(fedframework.MasterUpgrade(cluster.Name, k8sVersion))
 		framework.ExpectNoError(framework.CheckMasterVersion(cluster.Clientset, k8sVersion))
 
 		// TODO: Need to add Node upgrade. Add once this framework is stable
diff --git a/test/e2e_federation/util.go b/test/e2e_federation/util.go
index 21411ee017a..97743e2fd62 100644
--- a/test/e2e_federation/util.go
+++ b/test/e2e_federation/util.go
@@ -25,9 +25,6 @@ import (
 	"k8s.io/apimachinery/pkg/util/intstr"
 	"k8s.io/apimachinery/pkg/util/rand"
 	"k8s.io/apimachinery/pkg/util/wait"
-	restclient "k8s.io/client-go/rest"
-	"k8s.io/client-go/tools/clientcmd"
-	clientcmdapi "k8s.io/client-go/tools/clientcmd/api"
 	federationapi "k8s.io/kubernetes/federation/apis/federation/v1beta1"
 	fedclientset "k8s.io/kubernetes/federation/client/clientset_generated/federation_clientset"
 	"k8s.io/kubernetes/pkg/api"
@@ -42,18 +39,12 @@ import (
 )
 
 var (
-	KubeAPIQPS            float32 = 20.0
-	KubeAPIBurst                  = 30
-	DefaultFederationName         = "e2e-federation"
-	UserAgentName                 = "federation-e2e"
+	DefaultFederationName = "e2e-federation"
 	// We use this to decide how long to wait for our DNS probes to succeed.
 	DNSTTL = 180 * time.Second // TODO: make k8s.io/kubernetes/federation/pkg/federation-controller/service.minDnsTtl exported, and import it here.
 )
 
 const (
-	federatedDefaultTestTimeout  = 5 * time.Minute
-	federatedClustersWaitTimeout = 1 * time.Minute
-
 	// [30000, 32767] is the allowed default service nodeport range and our
 	// tests just use the defaults.
 	FederatedSvcNodePortFirst = 30000
@@ -62,15 +53,6 @@ const (
 
 var FederationSuite common.Suite
 
-// cluster keeps track of the assorted objects and state related to each cluster
-// in the federation
-type cluster struct {
-	name string
-	*kubeclientset.Clientset
-	namespaceCreated bool    // Did we need to create a new namespace in this cluster?  If so, we should delete it.
-	backendPod       *v1.Pod // The backend pod, if one's been created.
-}
-
 func createClusterObjectOrFail(f *fedframework.Framework, context *fedframework.E2EContext) {
 	framework.Logf("Creating cluster object: %s (%s, secret: %s)", context.Name, context.Cluster.Cluster.Server, context.Name)
 	cluster := federationapi.Cluster{
@@ -97,144 +79,6 @@ func createClusterObjectOrFail(f *fedframework.Framework, context *fedframework.
 	framework.Logf("Successfully created cluster object: %s (%s, secret: %s)", context.Name, context.Cluster.Cluster.Server, context.Name)
 }
 
-func clusterIsReadyOrFail(f *fedframework.Framework, context *fedframework.E2EContext) {
-	c, err := f.FederationClientset.Federation().Clusters().Get(context.Name, metav1.GetOptions{})
-	framework.ExpectNoError(err, fmt.Sprintf("get cluster: %+v", err))
-	if c.ObjectMeta.Name != context.Name {
-		framework.Failf("cluster name does not match input context: actual=%+v, expected=%+v", c, context)
-	}
-	err = isReady(context.Name, f.FederationClientset)
-	framework.ExpectNoError(err, fmt.Sprintf("unexpected error in verifying if cluster %s is ready: %+v", context.Name, err))
-	framework.Logf("Cluster %s is Ready", context.Name)
-}
-
-// waitForAllRegisteredClusters waits for all clusters defined in e2e context to be created
-// return ClusterList until the listed cluster items equals clusterCount
-func waitForAllRegisteredClusters(f *fedframework.Framework, clusterCount int) *federationapi.ClusterList {
-	var clusterList *federationapi.ClusterList
-	if err := wait.PollImmediate(framework.Poll, federatedClustersWaitTimeout, func() (bool, error) {
-		var err error
-		clusterList, err = f.FederationClientset.Federation().Clusters().List(metav1.ListOptions{})
-		if err != nil {
-			return false, err
-		}
-		framework.Logf("%d clusters registered, waiting for %d", len(clusterList.Items), clusterCount)
-		if len(clusterList.Items) == clusterCount {
-			return true, nil
-		}
-		return false, nil
-	}); err != nil {
-		framework.Failf("Failed to list registered clusters: %+v", err)
-	}
-	return clusterList
-}
-
-func createClientsetForCluster(c federationapi.Cluster, i int, userAgentName string) *kubeclientset.Clientset {
-	kubecfg, err := clientcmd.LoadFromFile(framework.TestContext.KubeConfig)
-	framework.ExpectNoError(err, "error loading KubeConfig: %v", err)
-
-	cfgOverride := &clientcmd.ConfigOverrides{
-		ClusterInfo: clientcmdapi.Cluster{
-			Server: c.Spec.ServerAddressByClientCIDRs[0].ServerAddress,
-		},
-	}
-	ccfg := clientcmd.NewNonInteractiveClientConfig(*kubecfg, c.Name, cfgOverride, clientcmd.NewDefaultClientConfigLoadingRules())
-	cfg, err := ccfg.ClientConfig()
-	framework.ExpectNoError(err, "Error creating client config in cluster #%d (%q)", i, c.Name)
-
-	cfg.QPS = KubeAPIQPS
-	cfg.Burst = KubeAPIBurst
-	return kubeclientset.NewForConfigOrDie(restclient.AddUserAgent(cfg, userAgentName))
-}
-
-// Creates the federation namespace in all underlying clusters.
-func createNamespaceInClusters(clusters map[string]*cluster, f *fedframework.Framework) {
-	nsName := f.FederationNamespace.Name
-	for name, c := range clusters {
-		// The e2e Framework created the required namespace in federation control plane, but we need to create it in all the others, if it doesn't yet exist.
-		// TODO(nikhiljindal): remove this once we have the namespace controller working as expected.
-		if _, err := c.Clientset.Core().Namespaces().Get(nsName, metav1.GetOptions{}); errors.IsNotFound(err) {
-			ns := &v1.Namespace{
-				ObjectMeta: metav1.ObjectMeta{
-					Name: nsName,
-				},
-			}
-			_, err := c.Clientset.Core().Namespaces().Create(ns)
-			if err == nil {
-				c.namespaceCreated = true
-			}
-			framework.ExpectNoError(err, "Couldn't create the namespace %s in cluster %q", nsName, name)
-			framework.Logf("Namespace %s created in cluster %q", nsName, name)
-		} else if err != nil {
-			framework.Logf("Couldn't create the namespace %s in cluster %q: %v", nsName, name, err)
-		}
-	}
-}
-
-// Unregisters the given clusters from federation control plane.
-// Also deletes the federation namespace from each cluster.
-func unregisterClusters(clusters map[string]*cluster, f *fedframework.Framework) {
-	nsName := f.FederationNamespace.Name
-	for name, c := range clusters {
-		if c.namespaceCreated {
-			if _, err := c.Clientset.Core().Namespaces().Get(nsName, metav1.GetOptions{}); !errors.IsNotFound(err) {
-				err := c.Clientset.Core().Namespaces().Delete(nsName, &metav1.DeleteOptions{})
-				framework.ExpectNoError(err, "Couldn't delete the namespace %s in cluster %q: %v", nsName, name, err)
-			}
-			framework.Logf("Namespace %s deleted in cluster %q", nsName, name)
-		}
-	}
-
-	// Delete the registered clusters in the federation API server.
-	clusterList, err := f.FederationClientset.Federation().Clusters().List(metav1.ListOptions{})
-	framework.ExpectNoError(err, "Error listing clusters")
-	for _, cluster := range clusterList.Items {
-		err := f.FederationClientset.Federation().Clusters().Delete(cluster.Name, &metav1.DeleteOptions{})
-		framework.ExpectNoError(err, "Error deleting cluster %q", cluster.Name)
-	}
-}
-
-// waitForNamespaceInFederatedClusters waits for the federated namespace to be created in federated clusters
-func waitForNamespaceInFederatedClusters(clusters map[string]*cluster, nsName string, timeout time.Duration) {
-	for name, c := range clusters {
-		err := wait.PollImmediate(framework.Poll, timeout, func() (bool, error) {
-			_, err := c.Clientset.Core().Namespaces().Get(nsName, metav1.GetOptions{})
-			if err != nil {
-				By(fmt.Sprintf("Waiting for namespace %q to be created in cluster %q, err: %v", nsName, name, err))
-				return false, nil
-			}
-			By(fmt.Sprintf("Namespace %q exists in cluster %q", nsName, name))
-			return true, nil
-		})
-		framework.ExpectNoError(err, "Failed to verify federated namespace %q creation in cluster %q", nsName, name)
-	}
-}
-
-// can not be moved to util, as By and Expect must be put in Ginkgo test unit
-func getRegisteredClusters(userAgentName string, f *fedframework.Framework) (map[string]*cluster, string) {
-	clusters := make(map[string]*cluster)
-	contexts := f.GetUnderlyingFederatedContexts()
-
-	By("Obtaining a list of all the clusters")
-	clusterList := waitForAllRegisteredClusters(f, len(contexts))
-
-	framework.Logf("Checking that %d clusters are Ready", len(contexts))
-	for _, context := range contexts {
-		clusterIsReadyOrFail(f, &context)
-	}
-	framework.Logf("%d clusters are Ready", len(contexts))
-
-	primaryClusterName := clusterList.Items[0].Name
-	By(fmt.Sprintf("Labeling %q as the first cluster", primaryClusterName))
-	for i, c := range clusterList.Items {
-		framework.Logf("Creating a clientset for the cluster %s", c.Name)
-		Expect(framework.TestContext.KubeConfig).ToNot(Equal(""), "KubeConfig must be specified to load clusters' client config")
-		clusters[c.Name] = &cluster{c.Name, createClientsetForCluster(c, i, userAgentName), false, nil}
-	}
-	waitForNamespaceInFederatedClusters(clusters, f.FederationNamespace.Name, federatedDefaultTestTimeout)
-	return clusters, primaryClusterName
-}
-
 // waitForServiceOrFail waits until a service is either present or absent in the cluster specified by clientset.
 // If the condition is not met within timout, it fails the calling test.
 func waitForServiceOrFail(clientset *kubeclientset.Clientset, namespace string, service *v1.Service, present bool, timeout time.Duration) {
@@ -259,10 +103,10 @@ func waitForServiceOrFail(clientset *kubeclientset.Clientset, namespace string,
 }
 
 // waitForServiceShardsOrFail waits for the service to appear in all clusters
-func waitForServiceShardsOrFail(namespace string, service *v1.Service, clusters map[string]*cluster) {
+func waitForServiceShardsOrFail(namespace string, service *v1.Service, clusters fedframework.ClusterSlice) {
 	framework.Logf("Waiting for service %q in %d clusters", service.Name, len(clusters))
 	for _, c := range clusters {
-		waitForServiceOrFail(c.Clientset, namespace, service, true, federatedDefaultTestTimeout)
+		waitForServiceOrFail(c.Clientset, namespace, service, true, fedframework.FederatedDefaultTestTimeout)
 	}
 }
 
@@ -317,7 +161,7 @@ func deleteServiceOrFail(clientset *fedclientset.Clientset, namespace string, se
 	err := clientset.Services(namespace).Delete(serviceName, &metav1.DeleteOptions{OrphanDependents: orphanDependents})
 	framework.ExpectNoError(err, "Error deleting service %q from namespace %q", serviceName, namespace)
 	// Wait for the service to be deleted.
-	err = wait.Poll(5*time.Second, federatedDefaultTestTimeout, func() (bool, error) {
+	err = wait.Poll(5*time.Second, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 		_, err := clientset.Core().Services(namespace).Get(serviceName, metav1.GetOptions{})
 		if err != nil && errors.IsNotFound(err) {
 			return true, nil
@@ -330,12 +174,13 @@ func deleteServiceOrFail(clientset *fedclientset.Clientset, namespace string, se
 	}
 }
 
-func cleanupServiceShardsAndProviderResources(namespace string, service *v1.Service, clusters map[string]*cluster) {
+func cleanupServiceShardsAndProviderResources(namespace string, service *v1.Service, clusters fedframework.ClusterSlice) {
 	framework.Logf("Deleting service %q in %d clusters", service.Name, len(clusters))
-	for name, c := range clusters {
+	for _, c := range clusters {
+		name := c.Name
 		var cSvc *v1.Service
 
-		err := wait.PollImmediate(framework.Poll, federatedDefaultTestTimeout, func() (bool, error) {
+		err := wait.PollImmediate(framework.Poll, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 			var err error
 			cSvc, err = c.Clientset.Services(namespace).Get(service.Name, metav1.GetOptions{})
 			if err != nil && !errors.IsNotFound(err) {
@@ -352,15 +197,15 @@ func cleanupServiceShardsAndProviderResources(namespace string, service *v1.Serv
 		})
 
 		if err != nil || cSvc == nil {
-			By(fmt.Sprintf("Failed to find service %q in namespace %q, in cluster %q in %s", service.Name, namespace, name, federatedDefaultTestTimeout))
+			By(fmt.Sprintf("Failed to find service %q in namespace %q, in cluster %q in %s", service.Name, namespace, name, fedframework.FederatedDefaultTestTimeout))
 			continue
 		}
 
-		err = cleanupServiceShard(c.Clientset, name, namespace, cSvc, federatedDefaultTestTimeout)
+		err = cleanupServiceShard(c.Clientset, name, namespace, cSvc, fedframework.FederatedDefaultTestTimeout)
 		if err != nil {
 			framework.Logf("Failed to delete service %q in namespace %q, in cluster %q: %v", service.Name, namespace, name, err)
 		}
-		err = cleanupServiceShardLoadBalancer(name, cSvc, federatedDefaultTestTimeout)
+		err = cleanupServiceShardLoadBalancer(name, cSvc, fedframework.FederatedDefaultTestTimeout)
 		if err != nil {
 			framework.Logf("Failed to delete cloud provider resources for service %q in namespace %q, in cluster %q", service.Name, namespace, name)
 		}
@@ -493,9 +338,12 @@ func discoverService(f *fedframework.Framework, name string, exists bool, podNam
 	}
 }
 
-// createBackendPodsOrFail creates one pod in each cluster, and returns the created pods (in the same order as clusterClientSets).
-// If creation of any pod fails, the test fails (possibly with a partially created set of pods). No retries are attempted.
-func createBackendPodsOrFail(clusters map[string]*cluster, namespace string, name string) {
+// BackendPodMap maps a cluster name to a backend pod created in that cluster
+type BackendPodMap map[string]*v1.Pod
+
+// createBackendPodsOrFail creates one pod in each cluster, and returns the created pods.  If creation of any pod fails,
+// the test fails (possibly with a partially created set of pods). No retries are attempted.
+func createBackendPodsOrFail(clusters fedframework.ClusterSlice, namespace string, name string) BackendPodMap {
 	pod := &v1.Pod{
 		ObjectMeta: metav1.ObjectMeta{
 			Name: name,
@@ -512,22 +360,24 @@ func createBackendPodsOrFail(clusters map[string]*cluster, namespace string, nam
 			RestartPolicy: v1.RestartPolicyAlways,
 		},
 	}
-	for name, c := range clusters {
+	podMap := make(BackendPodMap)
+	for _, c := range clusters {
+		name := c.Name
 		By(fmt.Sprintf("Creating pod %q in namespace %q in cluster %q", pod.Name, namespace, name))
 		createdPod, err := c.Clientset.Core().Pods(namespace).Create(pod)
 		framework.ExpectNoError(err, "Creating pod %q in namespace %q in cluster %q", name, namespace, name)
 		By(fmt.Sprintf("Successfully created pod %q in namespace %q in cluster %q: %v", pod.Name, namespace, name, *createdPod))
-		c.backendPod = createdPod
+		podMap[name] = createdPod
 	}
+	return podMap
 }
 
 // deleteOneBackendPodOrFail deletes exactly one backend pod which must not be nil
 // The test fails if there are any errors.
-func deleteOneBackendPodOrFail(c *cluster) {
-	pod := c.backendPod
+func deleteOneBackendPodOrFail(c *fedframework.Cluster, pod *v1.Pod) {
 	Expect(pod).ToNot(BeNil())
 	err := c.Clientset.Core().Pods(pod.Namespace).Delete(pod.Name, metav1.NewDeleteOptions(0))
-	msgFmt := fmt.Sprintf("Deleting Pod %q in namespace %q in cluster %q %%v", pod.Name, pod.Namespace, c.name)
+	msgFmt := fmt.Sprintf("Deleting Pod %q in namespace %q in cluster %q %%v", pod.Name, pod.Namespace, c.Name)
 	if errors.IsNotFound(err) {
 		framework.Logf(msgFmt, "does not exist. No need to delete it.")
 		return
@@ -538,13 +388,15 @@ func deleteOneBackendPodOrFail(c *cluster) {
 
 // deleteBackendPodsOrFail deletes one pod from each cluster that has one.
 // If deletion of any pod fails, the test fails (possibly with a partially deleted set of pods). No retries are attempted.
-func deleteBackendPodsOrFail(clusters map[string]*cluster, namespace string) {
-	for name, c := range clusters {
-		if c.backendPod != nil {
-			deleteOneBackendPodOrFail(c)
-			c.backendPod = nil
+func deleteBackendPodsOrFail(clusters fedframework.ClusterSlice, backendPods BackendPodMap) {
+	if backendPods == nil {
+		return
+	}
+	for _, c := range clusters {
+		if pod, ok := backendPods[c.Name]; ok {
+			deleteOneBackendPodOrFail(c, pod)
 		} else {
-			By(fmt.Sprintf("No backend pod to delete for cluster %q", name))
+			By(fmt.Sprintf("No backend pod to delete for cluster %q", c.Name))
 		}
 	}
 }
@@ -552,7 +404,7 @@ func deleteBackendPodsOrFail(clusters map[string]*cluster, namespace string) {
 // waitForReplicatSetToBeDeletedOrFail waits for the named ReplicaSet in namespace to be deleted.
 // If the deletion fails, the enclosing test fails.
 func waitForReplicaSetToBeDeletedOrFail(clientset *fedclientset.Clientset, namespace string, replicaSet string) {
-	err := wait.Poll(5*time.Second, federatedDefaultTestTimeout, func() (bool, error) {
+	err := wait.Poll(5*time.Second, fedframework.FederatedDefaultTestTimeout, func() (bool, error) {
 		_, err := clientset.Extensions().ReplicaSets(namespace).Get(replicaSet, metav1.GetOptions{})
 		if err != nil && errors.IsNotFound(err) {
 			return true, nil