Multi-Cloud戦略&マイグレーション完全ガイド2025：AWS/GCP/Azure比較、ハイブリッドクラウド

                    ┌─────────────────────┐
                    │   Global DNS / LB   │
                    │  (Route 53 / CF)    │
                    └──────────┬──────────┘
                         ┌─────┴─────┐
                         │           │
                    ┌────▼────┐ ┌────▼────┐
                    │  AWS    │ │  GCP    │
                    │ Region  │ │ Region  │
                    │         │ │         │
                    │ ┌─────┐ │ │ ┌─────┐ │
                    │ │ K8s │ │ │ │ GKE │ │
                    │ │ EKS │ │ │ │     │ │
                    │ └──┬──┘ │ │ └──┬──┘ │
                    │    │    │ │    │    │
                    │ ┌──▼──┐ │ │ ┌──▼──┐ │
                    │ │ DB  │◄┼─┼─► DB  │ │
                    │ │ RDS │ │ │ │ SQL │ │
                    │ └─────┘ │ │ └─────┘ │
                    └─────────┘ └─────────┘

                    ┌────────────────────┐
                    │    Global DNS      │
                    └────────┬───────────┘
                             │
                    ┌────────▼────────┐          ┌────────────────┐
                    │   AWS (Active)  │  複製 ──►│  Azure (Passive)│
                    │                 │          │                │
                    │  ┌───────────┐  │          │  ┌───────────┐ │
                    │  │ Workloads │  │          │  │ Standby   │ │
                    │  └───────────┘  │          │  └───────────┘ │
                    │  ┌───────────┐  │          │  ┌───────────┐ │
                    │  │ Database  │──┼──────────┼─►│ Replica   │ │
                    │  └───────────┘  │          │  └───────────┘ │
                    └─────────────────┘          └────────────────┘

# Kubernetes Federation - Cloud Bursting例
apiVersion: types.kubefed.io/v1beta1
kind: FederatedDeployment
metadata:
  name: web-app
  namespace: production
spec:
  template:
    spec:
      replicas: 10
      containers:
        - name: web
          image: registry.example.com/web-app:v2.1
  placement:
    clusters:
      - name: on-prem-cluster
        weight: 70
      - name: aws-eks-cluster
        weight: 20
      - name: gcp-gke-cluster
        weight: 10
  overrides:
    - clusterName: aws-eks-cluster
      clusterOverrides:
        - path: "/spec/replicas"
          value: 5

# クラウドコスト比較と自動配置の例
class CloudArbitrage:
    def __init__(self):
        self.providers = {
            "aws": AWSProvider(),
            "gcp": GCPProvider(),
            "azure": AzureProvider()
        }

    def find_optimal_placement(self, workload):
        costs = {}
        for name, provider in self.providers.items():
            cost = provider.estimate_cost(
                cpu=workload.cpu_cores,
                memory_gb=workload.memory_gb,
                storage_gb=workload.storage_gb,
                gpu=workload.gpu_type,
                duration_hours=workload.expected_duration,
                region=workload.preferred_region
            )
            costs[name] = cost

        # コスト対性能スコアを計算
        scores = {}
        for name, cost in costs.items():
            perf = self.providers[name].benchmark_score(workload.type)
            scores[name] = perf / cost

        best = max(scores, key=scores.get)
        return best, costs[best], scores[best]

    def auto_schedule(self, workloads):
        placements = []
        for wl in workloads:
            provider, cost, score = self.find_optimal_placement(wl)
            placements.append({
                "workload": wl.name,
                "provider": provider,
                "estimated_cost": cost,
                "efficiency_score": score
            })
        return placements

# Anthos Config Management - マルチクラスター設定
apiVersion: configmanagement.gke.io/v1
kind: ConfigManagement
metadata:
  name: config-management
spec:
  clusterName: production-cluster
  git:
    syncRepo: https://github.com/org/anthos-config
    syncBranch: main
    secretType: ssh
    policyDir: "policies"
  policyController:
    enabled: true
    templateLibraryInstalled: true
    referentialRulesEnabled: true
  hierarchyController:
    enabled: true
    enablePodTreeLabels: true
    enableHierarchicalResourceQuotas: true

# Azure ArcにKubernetesクラスターを登録
az connectedk8s connect \
  --name production-eks \
  --resource-group multi-cloud-rg \
  --location eastus \
  --tags "environment=production" "cloud=aws"

# Arc対応データサービスのデプロイ
az arcdata dc create \
  --name arc-dc \
  --k8s-namespace arc \
  --connectivity-mode indirect \
  --resource-group multi-cloud-rg \
  --location eastus \
  --storage-class managed-premium \
  --profile-name azure-arc-kubeadm

# Arc対応SQL Managed Instance
az sql mi-arc create \
  --name sql-prod \
  --resource-group multi-cloud-rg \
  --location eastus \
  --storage-class-data managed-premium \
  --storage-class-logs managed-premium \
  --cores-limit 8 \
  --memory-limit 32Gi \
  --k8s-namespace arc

# EKS Anywhereクラスター設定
apiVersion: anywhere.eks.amazonaws.com/v1alpha1
kind: Cluster
metadata:
  name: prod-cluster
spec:
  clusterNetwork:
    cniConfig:
      cilium: {}
    pods:
      cidrBlocks:
        - "192.168.0.0/16"
    services:
      cidrBlocks:
        - "10.96.0.0/12"
  controlPlaneConfiguration:
    count: 3
    endpoint:
      host: "10.0.0.100"
    machineGroupRef:
      kind: VSphereMachineConfig
      name: cp-machines
  workerNodeGroupConfigurations:
    - count: 5
      machineGroupRef:
        kind: VSphereMachineConfig
        name: worker-machines
      name: md-0
  kubernetesVersion: "1.29"
  managementCluster:
    name: prod-cluster

# OpenShiftマルチクラスターハブ設定
apiVersion: operator.open-cluster-management.io/v1
kind: MultiClusterHub
metadata:
  name: multiclusterhub
  namespace: open-cluster-management
spec:
  availabilityConfig: High
  enableClusterBackup: true
  overrides:
    components:
      - name: cluster-lifecycle
        enabled: true
      - name: cluster-backup
        enabled: true
      - name: multicluster-engine
        enabled: true
      - name: grc
        enabled: true
      - name: app-lifecycle
        enabled: true

ワークロード分析開始
       │
       ▼
  ビジネス価値がある？ ─── No ──► Retire（廃止）
       │ Yes
       ▼
  変更が必要？ ─── No ──► Retain（維持）
       │ Yes
       ▼
  SaaSで代替可能？ ─── Yes ──► Repurchase（再購入）
       │ No
       ▼
  アーキテクチャ変更が必要？ ─── No ──► Rehost（リホスト）
       │ Yes                              またはReplatform
       ▼
  完全な再設計が必要？ ─── No ──► Replatform（リプラットフォーム）
       │ Yes
       ▼
  Refactor（リファクター）

# マイグレーション戦略決定の自動化例
class MigrationAssessor:
    def assess_workload(self, workload):
        score = {
            "business_value": self._rate_business_value(workload),
            "technical_complexity": self._rate_complexity(workload),
            "cloud_readiness": self._rate_cloud_readiness(workload),
            "data_sensitivity": self._rate_data_sensitivity(workload),
            "dependency_count": len(workload.dependencies),
            "team_skill_level": self._rate_team_skills(workload.team)
        }

        # 戦略推奨ロジック
        if score["business_value"] < 3:
            return "Retire"
        if score["cloud_readiness"] < 2:
            return "Retain"
        if workload.has_saas_alternative and score["technical_complexity"] > 7:
            return "Repurchase"
        if score["technical_complexity"] < 4 and score["cloud_readiness"] > 6:
            return "Rehost"
        if score["technical_complexity"] < 7:
            return "Replatform"
        return "Refactor"

    def generate_report(self, workloads):
        results = {}
        for wl in workloads:
            strategy = self.assess_workload(wl)
            if strategy not in results:
                results[strategy] = []
            results[strategy].append({
                "name": wl.name,
                "estimated_effort_weeks": self._estimate_effort(wl, strategy),
                "estimated_cost": self._estimate_cost(wl, strategy),
                "risk_level": self._assess_risk(wl, strategy)
            })
        return results

# AWS Application Discovery Serviceの使用
aws discovery start-continuous-export

# エージェントベースの収集
aws discovery start-data-collection-by-agent-ids \
  --agent-ids agent-001 agent-002 agent-003

# サーバー依存関係マップの照会
aws discovery describe-agents \
  --filters name=hostName,values=web-server-*,condition=CONTAINS

# TCO比較分析フレームワーク
class TCOAnalysis:
    def calculate_on_prem_tco(self, infra):
        annual_costs = {
            "hardware": infra.server_count * 8000 / 3,  # 3年減価償却
            "software_licenses": infra.license_costs,
            "datacenter": infra.rack_units * 1200,  # 電力、冷却、スペース
            "network": infra.bandwidth_gbps * 500,
            "personnel": infra.fte_count * 120000,
            "maintenance": infra.server_count * 2400,
            "disaster_recovery": infra.dr_cost_annual,
            "security": infra.security_cost_annual,
            "compliance": infra.compliance_cost_annual
        }
        return sum(annual_costs.values()), annual_costs

    def calculate_cloud_tco(self, workloads, provider="aws"):
        annual_costs = {
            "compute": self._estimate_compute(workloads, provider),
            "storage": self._estimate_storage(workloads, provider),
            "network": self._estimate_network(workloads, provider),
            "managed_services": self._estimate_managed(workloads, provider),
            "personnel": workloads.cloud_fte * 130000,
            "migration_amortized": workloads.migration_cost / 3,
            "training": workloads.team_size * 5000,
            "tools": workloads.tool_licenses
        }
        return sum(annual_costs.values()), annual_costs

    def compare(self, infra, workloads):
        on_prem_total, on_prem_detail = self.calculate_on_prem_tco(infra)
        cloud_totals = {}
        for provider in ["aws", "gcp", "azure"]:
            total, detail = self.calculate_cloud_tco(workloads, provider)
            cloud_totals[provider] = {
                "total": total,
                "detail": detail,
                "savings_pct": (on_prem_total - total) / on_prem_total * 100
            }
        return {
            "on_prem": {"total": on_prem_total, "detail": on_prem_detail},
            "cloud": cloud_totals
        }

# AWS DMS（Database Migration Service）タスク設定例
# ソース：オンプレミスOracle、ターゲット：Amazon Aurora PostgreSQL
Resources:
  DMSReplicationTask:
    Type: AWS::DMS::ReplicationTask
    Properties:
      MigrationType: full-load-and-cdc
      SourceEndpointArn: !Ref OracleSourceEndpoint
      TargetEndpointArn: !Ref AuroraTargetEndpoint
      ReplicationInstanceArn: !Ref DMSReplicationInstance
      TableMappings: |
        {
          "rules": [
            {
              "rule-type": "selection",
              "rule-id": "1",
              "rule-name": "select-all-tables",
              "object-locator": {
                "schema-name": "PROD_SCHEMA",
                "table-name": "%"
              },
              "rule-action": "include"
            },
            {
              "rule-type": "transformation",
              "rule-id": "2",
              "rule-name": "lowercase-schema",
              "rule-action": "convert-lowercase",
              "rule-target": "schema",
              "object-locator": {
                "schema-name": "PROD_SCHEMA"
              }
            }
          ]
        }

# AWSからGCPへストレージ転送
gcloud transfer jobs create \
  --source-agent-pool=transfer-pool \
  --source-s3-bucket=my-aws-bucket \
  --source-s3-region=us-east-1 \
  --destination-gcs-bucket=my-gcp-bucket \
  --schedule-starts=2025-01-15T00:00:00Z \
  --schedule-repeats-every=24h \
  --include-prefixes="data/,backups/" \
  --exclude-prefixes="temp/,logs/"

Phase 1: プロキシレイヤー追加
┌────────────────┐
│ API Gateway /  │
│ Load Balancer  │
└───────┬────────┘
        │
        ▼
┌────────────────┐
│   Monolith     │  ← 全トラフィック
│ （オンプレミス） │
└────────────────┘

Phase 2: 一部機能を移行
┌────────────────┐
│ API Gateway    │
└───┬────────┬───┘
    │        │
    ▼        ▼
┌──────┐ ┌──────────┐
│ New  │ │ Monolith │  ← 残りのトラフィック
│ Auth │ │          │
│(Cloud)│ │(オンプレ) │
└──────┘ └──────────┘

Phase 3: 大部分の移行完了
┌────────────────┐
│ API Gateway    │
└┬──┬──┬──┬──┬───┘
 │  │  │  │  │
 ▼  ▼  ▼  ▼  ▼
┌─┐┌─┐┌─┐┌─┐┌────────┐
│A││B││C││D││Monolith│ ← 最小トラフィック
└─┘└─┘└─┘└─┘└────────┘
 (Cloud services)

# KubernetesベースのBlue-Green切り替え設定
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
  name: web-app-migration
spec:
  replicas: 10
  strategy:
    blueGreen:
      activeService: web-app-active
      previewService: web-app-preview
      autoPromotionEnabled: false
      prePromotionAnalysis:
        templates:
          - templateName: migration-validation
        args:
          - name: service-name
            value: web-app-preview
      postPromotionAnalysis:
        templates:
          - templateName: post-migration-check
      scaleDownDelaySeconds: 3600
  selector:
    matchLabels:
      app: web-app
  template:
    metadata:
      labels:
        app: web-app
        version: cloud-native
    spec:
      containers:
        - name: web-app
          image: registry.example.com/web-app:v3.0-cloud
          resources:
            requests:
              cpu: "500m"
              memory: "512Mi"
            limits:
              cpu: "1000m"
              memory: "1Gi"

┌─────────────────────────────────────────────────────┐
│              接続オプション比較                        │
├────────────────┬──────────┬──────────┬──────────────┤
│                │ 帯域幅    │ レイテンシ │  コスト      │
├────────────────┼──────────┼──────────┼──────────────┤
│ パブリックインターネット│ 変動    │ 高      │ エグレス費用 │
│ VPN (IPsec)    │ 1-3 Gbps│ 中      │ 低           │
│ 専用回線        │ 10-100Gb│ 低      │ 高（月額）    │
│ Megaport/Equinix│ 柔軟   │ 低      │ 中           │
└────────────────┴──────────┴──────────┴──────────────┘

# Terraform - AWS Transit GatewayとVPN設定
resource "aws_ec2_transit_gateway" "main" {
  description                     = "Multi-cloud transit gateway"
  default_route_table_association = "disable"
  default_route_table_propagation = "disable"
  auto_accept_shared_attachments  = "enable"

  tags = {
    Name = "multi-cloud-tgw"
  }
}

resource "aws_vpn_connection" "to_gcp" {
  customer_gateway_id = aws_customer_gateway.gcp.id
  transit_gateway_id  = aws_ec2_transit_gateway.main.id
  type                = "ipsec.1"
  static_routes_only  = false

  tunnel1_inside_cidr   = "169.254.10.0/30"
  tunnel2_inside_cidr   = "169.254.10.4/30"

  tags = {
    Name = "aws-to-gcp-vpn"
  }
}

resource "aws_customer_gateway" "gcp" {
  bgp_asn    = 65000
  ip_address = var.gcp_vpn_gateway_ip
  type       = "ipsec.1"

  tags = {
    Name = "gcp-customer-gateway"
  }
}

# Istioマルチクラスター設定（Primary-Remote）
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
metadata:
  name: istio-primary
spec:
  values:
    global:
      meshID: multi-cloud-mesh
      multiCluster:
        clusterName: aws-cluster
      network: aws-network
    pilot:
      env:
        EXTERNAL_ISTIOD: "true"
  meshConfig:
    defaultConfig:
      proxyMetadata:
        ISTIO_META_DNS_CAPTURE: "true"
        ISTIO_META_DNS_AUTO_ALLOCATE: "true"
  components:
    ingressGateways:
      - name: istio-eastwestgateway
        label:
          istio: eastwestgateway
          topology.istio.io/network: aws-network
        enabled: true
        k8s:
          env:
            - name: ISTIO_META_REQUESTED_NETWORK_VIEW
              value: aws-network

┌─────────────────────────────────────┐
│   中央アイデンティティプロバイダー     │
│    (Okta / Azure AD / Google)       │
└──────────┬──────────────────────────┘
           │ SAML / OIDC
     ┌─────┼──────┬───────┐
     ▼     ▼      ▼       ▼
  ┌─────┐┌─────┐┌──────┐┌─────────┐
  │ AWS ││ GCP ││Azure ││On-Prem  │
  │ IAM ││ IAM ││ AD   ││  LDAP   │
  └─────┘└─────┘└──────┘└─────────┘

# AWSからGCPリソースへのアクセス（Workload Identity Federation）
import google.auth
from google.auth import impersonated_credentials
import boto3

class CrossCloudAuth:
    def get_gcp_credentials_from_aws(self):
        # AWS STSから現在の認証情報を確認
        sts = boto3.client("sts")
        aws_identity = sts.get_caller_identity()

        # GCP Workload Identity Federationを使用
        credentials, project = google.auth.default(
            scopes=["https://www.googleapis.com/auth/cloud-platform"]
        )

        # サービスアカウントのインパーソネーション
        target_credentials = impersonated_credentials.Credentials(
            source_credentials=credentials,
            target_principal="cross-cloud@project.iam.gserviceaccount.com",
            target_scopes=["https://www.googleapis.com/auth/cloud-platform"],
            lifetime=3600
        )

        return target_credentials

# マルチクラウドKubernetesデプロイ（Helm values）
# values-aws.yaml
cloud:
  provider: aws
  region: us-east-1
  storageClass: gp3
  ingressClass: alb
  serviceAnnotations:
    service.beta.kubernetes.io/aws-load-balancer-type: nlb
    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facing

# values-gcp.yaml
cloud:
  provider: gcp
  region: us-central1
  storageClass: pd-ssd
  ingressClass: gce
  serviceAnnotations:
    cloud.google.com/neg: '{"ingress": true}'

# values-azure.yaml
cloud:
  provider: azure
  region: eastus
  storageClass: managed-premium
  ingressClass: azure-application-gateway
  serviceAnnotations:
    service.beta.kubernetes.io/azure-load-balancer-internal: "false"

# modules/compute/main.tf - クラウド抽象化レイヤー
variable "cloud_provider" {
  type = string
  validation {
    condition     = contains(["aws", "gcp", "azure"], var.cloud_provider)
    error_message = "Supported providers: aws, gcp, azure"
  }
}

variable "instance_config" {
  type = object({
    name          = string
    cpu           = number
    memory_gb     = number
    disk_gb       = number
    os            = string
  })
}

# AWS実装
module "aws_compute" {
  source = "./aws"
  count  = var.cloud_provider == "aws" ? 1 : 0

  instance_type = local.aws_instance_map[
    "${var.instance_config.cpu}-${var.instance_config.memory_gb}"
  ]
  ami_id      = local.aws_ami_map[var.instance_config.os]
  volume_size = var.instance_config.disk_gb
  name        = var.instance_config.name
}

# GCP実装
module "gcp_compute" {
  source = "./gcp"
  count  = var.cloud_provider == "gcp" ? 1 : 0

  machine_type = local.gcp_machine_map[
    "${var.instance_config.cpu}-${var.instance_config.memory_gb}"
  ]
  image      = local.gcp_image_map[var.instance_config.os]
  disk_size  = var.instance_config.disk_gb
  name       = var.instance_config.name
}

# Azure実装
module "azure_compute" {
  source = "./azure"
  count  = var.cloud_provider == "azure" ? 1 : 0

  vm_size     = local.azure_vm_map[
    "${var.instance_config.cpu}-${var.instance_config.memory_gb}"
  ]
  image_ref   = local.azure_image_map[var.instance_config.os]
  disk_size   = var.instance_config.disk_gb
  name        = var.instance_config.name
}

output "instance_id" {
  value = coalesce(
    try(module.aws_compute[0].instance_id, ""),
    try(module.gcp_compute[0].instance_id, ""),
    try(module.azure_compute[0].instance_id, "")
  )
}

# マルチステージビルド - クラウド非依存イメージ
FROM golang:1.22-alpine AS builder

WORKDIR /app
COPY go.mod go.sum ./
RUN go mod download

COPY . .
RUN CGO_ENABLED=0 GOOS=linux go build -o /app/server ./cmd/server

# 最終イメージ - distroless（クラウド非依存）
FROM gcr.io/distroless/static-debian12:nonroot

COPY --from=builder /app/server /server
COPY --from=builder /app/config /config

EXPOSE 8080
USER nonroot:nonroot
ENTRYPOINT ["/server"]

# マルチクラウドDRオーケストレーター
class MultiCloudDR:
    def __init__(self):
        self.primary = AWSProvider(region="us-east-1")
        self.secondary = AzureProvider(region="eastus")
        self.health_checker = HealthChecker()

    def execute_failover(self):
        """Primary（AWS）障害時にSecondary（Azure）へ切り替え"""
        steps = [
            self._verify_secondary_health,
            self._promote_database_replica,
            self._update_dns_records,
            self._scale_up_secondary,
            self._verify_application_health,
            self._notify_stakeholders
        ]

        for step in steps:
            result = step()
            if not result.success:
                self._rollback_failover(result.step_index)
                raise FailoverError(f"Failed at step: {step.__name__}")

    def _promote_database_replica(self):
        """Azure SQL読み取りレプリカをプライマリに昇格"""
        self.secondary.promote_replica(
            server="dr-sql-server",
            database="production-db",
            failover_group="prod-failover-group"
        )

    def continuous_replication(self):
        """継続的なデータ同期"""
        replication_config = {
            "database": {
                "type": "async",
                "lag_threshold_seconds": 30,
                "source": "aws-rds-primary",
                "target": "azure-sql-replica"
            },
            "storage": {
                "type": "incremental",
                "interval_minutes": 15,
                "source": "s3://prod-bucket",
                "target": "azure://prod-container"
            }
        }
        return replication_config

# マルチクラウドコストダッシュボードコレクター
class MultiCloudCostCollector:
    def collect_all_costs(self, period="monthly"):
        aws_costs = self._get_aws_costs(period)
        gcp_costs = self._get_gcp_costs(period)
        azure_costs = self._get_azure_costs(period)

        return {
            "total": aws_costs["total"] + gcp_costs["total"] + azure_costs["total"],
            "by_provider": {
                "aws": aws_costs,
                "gcp": gcp_costs,
                "azure": azure_costs
            },
            "by_service": self._aggregate_by_service(
                aws_costs, gcp_costs, azure_costs
            ),
            "anomalies": self._detect_anomalies(
                aws_costs, gcp_costs, azure_costs
            ),
            "recommendations": self._generate_optimization_recommendations(
                aws_costs, gcp_costs, azure_costs
            )
        }

    def _detect_anomalies(self, *provider_costs):
        """コスト異常検知"""
        anomalies = []
        for costs in provider_costs:
            for service, cost in costs["by_service"].items():
                avg = cost.get("rolling_avg_30d", 0)
                current = cost.get("current", 0)
                if avg > 0 and current > avg * 1.5:
                    anomalies.append({
                        "service": service,
                        "provider": costs["provider"],
                        "current": current,
                        "average": avg,
                        "increase_pct": (current - avg) / avg * 100
                    })
        return anomalies

# Open Policy Agent (OPA) - マルチクラウドポリシー
package multicloud.governance

# 全リソースに必須タグを要求
required_tags := ["environment", "team", "cost-center", "data-classification"]

deny[msg] {
  resource := input.resource
  tag := required_tags[_]
  not resource.tags[tag]
  msg := sprintf(
    "Resource %v is missing required tag: %v",
    [resource.name, tag]
  )
}

# データ主権 - 特定データは特定リージョンのみに保存
deny[msg] {
  resource := input.resource
  resource.tags["data-classification"] == "pii-jp"
  not startswith(resource.region, "ap-northeast-1")
  not startswith(resource.region, "japan")
  msg := sprintf(
    "Japanese PII data must be stored in Japan region. Resource %v in %v",
    [resource.name, resource.region]
  )
}

# 暗号化必須
deny[msg] {
  resource := input.resource
  resource.type == "storage_bucket"
  not resource.encryption.enabled
  msg := sprintf(
    "Storage %v must have encryption enabled",
    [resource.name]
  )
}