chore(topograph): rename backend switch tiers

charts/topograph/templates/deployment.yaml

@@ -39,9 +39,9 @@ spec:
           args:
             - -v={{ .Values.verbosity }}
             - -k8s-topology-key-accelerator={{ .Values.topologyNodeLabels.accelerator }}
-            - -k8s-topology-key-block={{ .Values.topologyNodeLabels.block }}
+            - -k8s-topology-key-leaf={{ .Values.topologyNodeLabels.leaf }}
             - -k8s-topology-key-spine={{ .Values.topologyNodeLabels.spine }}
-            - -k8s-topology-key-datacenter={{ .Values.topologyNodeLabels.datacenter }}
+            - -k8s-topology-key-core={{ .Values.topologyNodeLabels.core }}
           env:
             - name: NODE_DATA_BROKER_NAME
               value: {{ printf "%s-%s" .Release.Name "node-data-broker" | trunc 63 | trimSuffix "-" }}

charts/topograph/values.yaml

@@ -47,9 +47,9 @@ config:
 
 topologyNodeLabels:
   accelerator: network.topology.nvidia.com/accelerator
-  block: network.topology.nvidia.com/block
+  leaf: network.topology.nvidia.com/leaf
   spine: network.topology.nvidia.com/spine
-  datacenter: network.topology.nvidia.com/datacenter
+  core: network.topology.nvidia.com/core
 
 podAnnotations: {}
 podLabels: {}

cmd/topograph/main.go

@@ -35,13 +35,13 @@ var GitTag string
 
 func main() {
 	var cfg string
-	var labelAccelerator, labelBlock, labelSpine, labelDatacenter string
+	var labelAccelerator, labelLeaf, labelSpine, labelCore string
 	var version bool
 	flag.StringVar(&cfg, "c", "/etc/topograph/topograph-config.yaml", "config file")
 	flag.StringVar(&labelAccelerator, "k8s-topology-key-accelerator", k8s.DefaultLabelAccelerator, "K8s node label for accelerated network type")
-	flag.StringVar(&labelBlock, "k8s-topology-key-block", k8s.DefaultLabelBlock, "K8s node label for the cluster's lower network tier")
+	flag.StringVar(&labelLeaf, "k8s-topology-key-leaf", k8s.DefaultLabelLeaf, "K8s node label for the cluster's lower network tier")
 	flag.StringVar(&labelSpine, "k8s-topology-key-spine", k8s.DefaultLabelSpine, "K8s node label for the cluster's middle network tier")
-	flag.StringVar(&labelDatacenter, "k8s-topology-key-datacenter", k8s.DefaultLabelDatacenter, "K8s node label for the cluster's top network tier")
+	flag.StringVar(&labelCore, "k8s-topology-key-core", k8s.DefaultLabelCore, "K8s node label for the cluster's top network tier")
 	flag.BoolVar(&version, "version", false, "show the version")
 
 	klog.InitFlags(nil)
@@ -53,7 +53,7 @@ func main() {
 		os.Exit(0)
 	}
 
-	k8s.InitLabels(labelAccelerator, labelBlock, labelSpine, labelDatacenter)
+	k8s.InitLabels(labelAccelerator, labelLeaf, labelSpine, labelCore)
 
 	if err := mainInternal(cfg); err != nil {
 		klog.Error(err.Error())

docs/engines/k8s.md

@@ -7,19 +7,19 @@ Topograph is a tool designed to enhance scheduling decisions in Kubernetes clust
 Topograph maps both the multi-tier network hierarchy and accelerated network domains (such as NVLink) using node labels.
 Most cloud providers expose three levels of network topology through their APIs. To provide a unified view, Topograph assigns four labels to each node:
 * `network.topology.nvidia.com/accelerator`: Identifies high-speed interconnect domains, such as NVLink.
-* `network.topology.nvidia.com/block`: Indicates the switches directly connected to compute nodes.
-* `network.topology.nvidia.com/spine`: Represents the next tier of switches above the block level.
-* `network.topology.nvidia.com/datacenter`: Denotes the top-level switches.
+* `network.topology.nvidia.com/leaf`: Indicates the switches directly connected to compute nodes.
+* `network.topology.nvidia.com/spine`: Represents the next tier of switches above the leaf level.
+* `network.topology.nvidia.com/core`: Denotes the top-level switches.
 
 The names of these node labels are configurable via the [Helm chart](https://github.com/NVIDIA/topograph/tree/main/charts/topograph).
 
 For example, if a node belongs to NVLink domain `nvl1` and connects to switch `s1`, which connects to switch `s2`, and then to switch `s3`, Topograph will apply the following labels to the node:
 
 ```
   network.topology.nvidia.com/accelerator: nvl1
-  network.topology.nvidia.com/block: s1
+  network.topology.nvidia.com/leaf: s1
   network.topology.nvidia.com/spine: s2
-  network.topology.nvidia.com/datacenter: s3
+  network.topology.nvidia.com/core: s3
 ```
 
 <p align="center"><img src="../assets/topograph-k8s.png" width="600" alt="Design"></p>
@@ -55,15 +55,15 @@ closer network proximity.
                     operator: In
                     values:
                       - myapp
-              topologyKey: network.topology.nvidia.com/block
+              topologyKey: network.topology.nvidia.com/leaf
 ```
-Pods are prioritized to be placed on nodes sharing the label `network.topology.nvidia.com/block`.
+Pods are prioritized to be placed on nodes sharing the label `network.topology.nvidia.com/leaf`.
 These nodes are connected to the same network switch, ensuring the lowest latency for communication.
 
 Nodes with the label `network.topology.nvidia.com/spine` are next in priority.
 Pods on these nodes will still be relatively close, but with slightly higher latency.
 
-In the three-tier network, all nodes will share the same `network.topology.nvidia.com/datacenter` label,
+In the three-tier network, all nodes will share the same `network.topology.nvidia.com/core` label,
 so it doesn’t need to be included in pod affinity settings.
 
 Since the default Kubernetes scheduler places one pod at a time, the placement may vary depending on where

pkg/engines/k8s/labeler.go

@@ -26,23 +26,23 @@ import (
 
 const (
 	DefaultLabelAccelerator = "network.topology.nvidia.com/accelerator"
-	DefaultLabelBlock       = "network.topology.nvidia.com/block"
+	DefaultLabelLeaf        = "network.topology.nvidia.com/leaf"
 	DefaultLabelSpine       = "network.topology.nvidia.com/spine"
-	DefaultLabelDatacenter  = "network.topology.nvidia.com/datacenter"
+	DefaultLabelCore        = "network.topology.nvidia.com/core"
 )
 
 var (
-	labelAccelerator, labelBlock, labelSpine, labelDatacenter string
+	labelAccelerator, labelLeaf, labelSpine, labelCore string
 
 	switchNetworkHierarchy []string
 )
 
-func InitLabels(accelerator, block, spine, datacenter string) {
+func InitLabels(accelerator, leaf, spine, core string) {
 	labelAccelerator = accelerator
-	labelBlock = block
+	labelLeaf = leaf
 	labelSpine = spine
-	labelDatacenter = datacenter
-	switchNetworkHierarchy = []string{labelBlock, labelSpine, labelDatacenter}
+	labelCore = core
+	switchNetworkHierarchy = []string{labelLeaf, labelSpine, labelCore}
 }
 
 // map nodename:[label name: label value]

pkg/engines/k8s/labeler_test.go

@@ -39,16 +39,16 @@ func (l *testLabeler) AddNodeLabels(_ context.Context, nodeName string, labels m
 }
 
 func TestApplyNodeLabelsWithTree(t *testing.T) {
-	InitLabels(DefaultLabelAccelerator, DefaultLabelBlock, DefaultLabelSpine, DefaultLabelDatacenter)
+	InitLabels(DefaultLabelAccelerator, DefaultLabelLeaf, DefaultLabelSpine, DefaultLabelCore)
 	root, _ := translate.GetTreeTestSet(true)
 	labeler := &testLabeler{data: make(map[string]map[string]string)}
 	data := map[string]map[string]string{
-		"Node201": {"network.topology.nvidia.com/block": "S2", "network.topology.nvidia.com/spine": "S1"},
-		"Node202": {"network.topology.nvidia.com/block": "S2", "network.topology.nvidia.com/spine": "S1"},
-		"Node205": {"network.topology.nvidia.com/block": "S2", "network.topology.nvidia.com/spine": "S1"},
-		"Node304": {"network.topology.nvidia.com/block": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
-		"Node305": {"network.topology.nvidia.com/block": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
-		"Node306": {"network.topology.nvidia.com/block": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
+		"Node201": {"network.topology.nvidia.com/leaf": "S2", "network.topology.nvidia.com/spine": "S1"},
+		"Node202": {"network.topology.nvidia.com/leaf": "S2", "network.topology.nvidia.com/spine": "S1"},
+		"Node205": {"network.topology.nvidia.com/leaf": "S2", "network.topology.nvidia.com/spine": "S1"},
+		"Node304": {"network.topology.nvidia.com/leaf": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
+		"Node305": {"network.topology.nvidia.com/leaf": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
+		"Node306": {"network.topology.nvidia.com/leaf": "xf946c4acef2d5939", "network.topology.nvidia.com/spine": "S1"},
 	}
 
 	err := NewTopologyLabeler().ApplyNodeLabels(context.TODO(), root, labeler)
@@ -57,81 +57,81 @@ func TestApplyNodeLabelsWithTree(t *testing.T) {
 }
 
 func TestApplyNodeLabelsWithBlock(t *testing.T) {
-	InitLabels(DefaultLabelAccelerator, DefaultLabelBlock, DefaultLabelSpine, DefaultLabelDatacenter)
+	InitLabels(DefaultLabelAccelerator, DefaultLabelLeaf, DefaultLabelSpine, DefaultLabelCore)
 	root, _ := translate.GetBlockWithMultiIBTestSet()
 	labeler := &testLabeler{data: make(map[string]map[string]string)}
 	data := map[string]map[string]string{
 		"Node104": {
 			"network.topology.nvidia.com/accelerator": "B1",
-			"network.topology.nvidia.com/block":       "S2",
+			"network.topology.nvidia.com/leaf":        "S2",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node105": {
 			"network.topology.nvidia.com/accelerator": "B1",
-			"network.topology.nvidia.com/block":       "S2",
+			"network.topology.nvidia.com/leaf":        "S2",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node106": {
 			"network.topology.nvidia.com/accelerator": "B1",
-			"network.topology.nvidia.com/block":       "S2",
+			"network.topology.nvidia.com/leaf":        "S2",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node201": {
 			"network.topology.nvidia.com/accelerator": "B2",
-			"network.topology.nvidia.com/block":       "S3",
+			"network.topology.nvidia.com/leaf":        "S3",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node202": {
 			"network.topology.nvidia.com/accelerator": "B2",
-			"network.topology.nvidia.com/block":       "S3",
+			"network.topology.nvidia.com/leaf":        "S3",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node205": {
 			"network.topology.nvidia.com/accelerator": "B2",
-			"network.topology.nvidia.com/block":       "S3",
+			"network.topology.nvidia.com/leaf":        "S3",
 			"network.topology.nvidia.com/spine":       "S1",
-			"network.topology.nvidia.com/datacenter":  "IB2",
+			"network.topology.nvidia.com/core":        "IB2",
 		},
 		"Node301": {
 			"network.topology.nvidia.com/accelerator": "B3",
-			"network.topology.nvidia.com/block":       "S5",
+			"network.topology.nvidia.com/leaf":        "S5",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 		"Node302": {
 			"network.topology.nvidia.com/accelerator": "B3",
-			"network.topology.nvidia.com/block":       "S5",
+			"network.topology.nvidia.com/leaf":        "S5",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 		"Node303": {
 			"network.topology.nvidia.com/accelerator": "B3",
-			"network.topology.nvidia.com/block":       "S5",
+			"network.topology.nvidia.com/leaf":        "S5",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 		"Node401": {
 			"network.topology.nvidia.com/accelerator": "B4",
-			"network.topology.nvidia.com/block":       "S6",
+			"network.topology.nvidia.com/leaf":        "S6",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 		"Node402": {
 			"network.topology.nvidia.com/accelerator": "B4",
-			"network.topology.nvidia.com/block":       "S6",
+			"network.topology.nvidia.com/leaf":        "S6",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 		"Node403": {
 			"network.topology.nvidia.com/accelerator": "B4",
-			"network.topology.nvidia.com/block":       "S6",
+			"network.topology.nvidia.com/leaf":        "S6",
 			"network.topology.nvidia.com/spine":       "S4",
-			"network.topology.nvidia.com/datacenter":  "IB1",
+			"network.topology.nvidia.com/core":        "IB1",
 		},
 	}
 

pkg/providers/aws/instance_topology.go

@@ -102,10 +102,10 @@ func (p *baseProvider) generateRegionInstanceTopology(ctx context.Context, pageS
 
 func convert(inst *types.InstanceTopology) *topology.InstanceTopology {
 	topo := &topology.InstanceTopology{
-		InstanceID:   *inst.InstanceId,
-		BlockID:      inst.NetworkNodes[2],
-		SpineID:      inst.NetworkNodes[1],
-		DatacenterID: inst.NetworkNodes[0],
+		InstanceID: *inst.InstanceId,
+		LeafID:     inst.NetworkNodes[2],
+		SpineID:    inst.NetworkNodes[1],
+		CoreID:     inst.NetworkNodes[0],
 	}
 	if inst.CapacityBlockId != nil {
 		topo.AcceleratorID = *inst.CapacityBlockId

pkg/providers/gcp/instance_topology.go

@@ -101,12 +101,12 @@ func (p *baseProvider) generateRegionInstanceTopology(ctx context.Context, clien
 					continue
 				}
 				inst := &topology.InstanceTopology{
-					InstanceID:   instanceId,
-					DatacenterID: instance.ResourceStatus.PhysicalHostTopology.GetCluster(),
-					SpineID:      instance.ResourceStatus.PhysicalHostTopology.GetBlock(),
-					BlockID:      instance.ResourceStatus.PhysicalHostTopology.GetSubblock(),
+					InstanceID: instanceId,
+					CoreID:     instance.ResourceStatus.PhysicalHostTopology.GetCluster(),
+					SpineID:    instance.ResourceStatus.PhysicalHostTopology.GetBlock(),
+					LeafID:     instance.ResourceStatus.PhysicalHostTopology.GetSubblock(),
 				}
-				inst.AcceleratorID = inst.BlockID
+				inst.AcceleratorID = inst.LeafID
 				klog.Infof("Adding topology: %s", inst.String())
 				topo.Append(inst)
 			}

pkg/providers/lambdai/instance_topology.go

@@ -55,11 +55,11 @@ func (p *baseProvider) generateRegionInstanceTopology(ctx context.Context, clien
 			for indx := range len(inst.NetworkPath) {
 				switch indx {
 				case 0:
-					t.BlockID = inst.NetworkPath[indx]
+					t.LeafID = inst.NetworkPath[indx]
 				case 1:
 					t.SpineID = inst.NetworkPath[indx]
 				case 2:
-					t.DatacenterID = inst.NetworkPath[indx]
+					t.CoreID = inst.NetworkPath[indx]
 				default:
 					klog.Warningf("unsupported size %d of topology path for instance %q", len(inst.NetworkPath), inst.ID)
 				}

pkg/providers/nebius/instance_topology.go

@@ -71,9 +71,9 @@ func (p *baseProvider) generateRegionInstanceTopology(ctx context.Context, clien
 			path := ibTopology.GetPath()
 			switch len(path) {
 			case 3:
-				inst.DatacenterID = path[0]
+				inst.CoreID = path[0]
 				inst.SpineID = path[1]
-				inst.BlockID = path[2]
+				inst.LeafID = path[2]
 			default:
 				klog.Warningf("unsupported size %d of topology path for node %q", len(path), hostname)
 				continue

pkg/providers/oci/instance_topology.go

@@ -87,10 +87,10 @@ func convert(host *core.ComputeHostSummary) (*topology.InstanceTopology, error)
 	}
 
 	topo := &topology.InstanceTopology{
-		InstanceID:   *host.InstanceId,
-		BlockID:      *host.LocalBlockId,
-		SpineID:      *host.NetworkBlockId,
-		DatacenterID: *host.HpcIslandId,
+		InstanceID: *host.InstanceId,
+		LeafID:     *host.LocalBlockId,
+		SpineID:    *host.NetworkBlockId,
+		CoreID:     *host.HpcIslandId,
 	}
 
 	if host.GpuMemoryFabricId != nil {

pkg/providers/oci/instance_topology_test.go

@@ -27,10 +27,10 @@ import (
 
 func TestConvert(t *testing.T) {
 	valid := &topology.InstanceTopology{
-		InstanceID:   "id",
-		BlockID:      "block",
-		SpineID:      "net",
-		DatacenterID: "datacenter",
+		InstanceID: "id",
+		LeafID:     "leaf",
+		SpineID:    "net",
+		CoreID:     "core",
 	}
 
 	testCases := []struct {
@@ -55,15 +55,15 @@ func TestConvert(t *testing.T) {
 			name: "Case 3: missing NetworkBlockId",
 			host: &core.ComputeHostSummary{
 				InstanceId:   &valid.InstanceID,
-				LocalBlockId: &valid.BlockID,
+				LocalBlockId: &valid.LeafID,
 			},
 			err: `missing NetworkBlockId for instance "id"`,
 		},
 		{
 			name: "Case 4: missing HpcIslandId",
 			host: &core.ComputeHostSummary{
 				InstanceId:     &valid.InstanceID,
-				LocalBlockId:   &valid.BlockID,
+				LocalBlockId:   &valid.LeafID,
 				NetworkBlockId: &valid.SpineID,
 			},
 			err: `missing HpcIslandId for instance "id"`,
@@ -72,9 +72,9 @@ func TestConvert(t *testing.T) {
 			name: "Case 5: valid input",
 			host: &core.ComputeHostSummary{
 				InstanceId:     &valid.InstanceID,
-				LocalBlockId:   &valid.BlockID,
+				LocalBlockId:   &valid.LeafID,
 				NetworkBlockId: &valid.SpineID,
-				HpcIslandId:    &valid.DatacenterID,
+				HpcIslandId:    &valid.CoreID,
 			},
 			topo: valid,
 		},

pkg/providers/oci/provider_imds.go

@@ -78,9 +78,9 @@ func (p *imdsProvider) getComputeHostInfo(ctx context.Context, ci topology.Compu
 		if nodeTopology, ok := topoMap[node]; ok {
 			topo.Instances = append(topo.Instances, &topology.InstanceTopology{
 				InstanceID:    instanceID,
-				BlockID:       nodeTopology.LocalBlock,
+				LeafID:        nodeTopology.LocalBlock,
 				SpineID:       nodeTopology.NetworkBlock,
-				DatacenterID:  nodeTopology.HPCIslandId,
+				CoreID:        nodeTopology.HPCIslandId,
 				AcceleratorID: nodeTopology.GpuMemoryFabric,
 			})
 		}

pkg/server/grpc_client.go

@@ -69,9 +69,9 @@ func forwardRequest(ctx context.Context, tr *topology.Request, url string, cis [
 func convert(inst *pb.Instance) *topology.InstanceTopology {
 	topo := &topology.InstanceTopology{
 		InstanceID:    inst.Id,
-		BlockID:       inst.NetworkLayers[0],
+		LeafID:        inst.NetworkLayers[0],
 		SpineID:       inst.NetworkLayers[1],
-		DatacenterID:  inst.NetworkLayers[2],
+		CoreID:        inst.NetworkLayers[2],
 		AcceleratorID: inst.NvlinkDomain,
 	}
 	klog.V(4).Infof("Adding instance topology %s", topo.String())