Multi-Cloud Strategy: Abstraction, Portability, Cost

Difficulty: Staff Level | Companies: Netflix, Google, Microsoft, HashiCorp, Pulumi

Interview Question

"Design a multi-cloud strategy for a global enterprise using AWS, Azure, and GCP. How do you handle abstraction, portability, cost optimization, and governance?"

ℹ️Key Concepts

This question tests your understanding of multi-cloud architecture, cloud abstraction, and enterprise governance.

Complete Multi-Cloud Architecture

Architecture Overview

Architecture Diagram

┌─────────────────────────────────────────────────────────────────────────┐
│                    MULTI-CLOUD ARCHITECTURE                              │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  ┌───────────────── ABSTRACTION LAYER ───────────────┐                 │
│  │                                                       │              │
│  │  ┌─────────────────────────────────────────────┐    │              │
│  │  │           Cloud Abstraction Layer             │    │              │
│  │  │  (Terraform / Pulumi / Crossplane)           │    │              │
│  │  └─────────────────────────────────────────────┘    │              │
│  │                                                       │              │
│  │  ┌─────────────────────────────────────────────┐    │              │
│  │  │           Kubernetes (EKS/AKS/GKE)           │    │              │
│  │  │  (Container Orchestration Abstraction)       │    │              │
│  │  └─────────────────────────────────────────────┘    │              │
│  │                                                       │              │
│  └──────────────────────┬──────────────────────────────┘              │
│                         │                                               │
│  ┌───────────────── CLOUD PROVIDERS ─────────────────┐                │
│  │                                                       │              │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────┐          │              │
│  │  │   AWS    │  │  Azure   │  │   GCP    │          │              │
│  │  │          │  │          │  │          │          │              │
│  │  │ ┌──────┐ │  │ ┌──────┐ │  │ ┌──────┐ │          │              │
│  │  │ │EKS   │ │  │ │AKS   │ │  │ │GKE   │ │          │              │
│  │  │ │RDS   │ │  │ │SQL DB│ │  │ │Cloud │ │          │              │
│  │  │ │S3    │ │  │ │Blob  │ │  │ │SQL   │ │          │              │
│  │  │ └──────┘ │  │ └──────┘ │  │ └──────┘ │          │              │
│  │  └──────────┘  └──────────┘  └──────────┘          │              │
│  │                                                       │              │
│  └──────────────────────┬──────────────────────────────┘              │
│                         │                                               │
│  ┌───────────────── GOVERNANCE LAYER ────────────────┐                │
│  │  Cost Management │ Security │ Compliance │ IAM     │                 │
│  └─────────────────────────────────────────────────────┘              │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

Mathematical Foundation: Multi-Cloud Metrics

Cloud Distribution:

Total workloads: W = 100
AWS workloads: A = 40%
Azure workloads: Z = 35%
GCP workloads: G = 25%

Cost Optimization:

AWS monthly cost: C_aws = $500,000
Azure monthly cost: C_azure = $450,000
GCP monthly cost: C_gcp = $400,000
Total monthly cost: C_total = $1,350,000
Optimized distribution savings: S = 15% = $202,500/month

Availability Calculation:

AWS availability: A_aws = 99.99%
Azure availability: A_azure = 99.99%
GCP availability: A_gcp = 99.99%
Multi-cloud availability: A_multi = 1 - (1 - A_aws) × (1 - A_azure) × (1 - A_gcp)
A_multi = 99.999999%

Latency Optimization:

AWS latency: L_aws = 50ms
Azure latency: L_azure = 45ms
GCP latency: L_gcp = 55ms
Best latency: L_best = min(L_aws, L_azure, L_gcp) = 45ms

Terraform Multi-Cloud Implementation

# Multi-cloud provider configuration
terraform {
  required_providers {
    aws = {
      source  = "hashicorp/aws"
      version = "~> 5.0"
    }
    azurerm = {
      source  = "hashicorp/azurerm"
      version = "~> 3.0"
    }
    google = {
      source  = "hashicorp/google"
      version = "~> 5.0"
    }
  }
}

provider "aws" {
  alias  = "us_east"
  region = "us-east-1"
}

provider "aws" {
  alias  = "eu_west"
  region = "eu-west-1"
}

provider "azurerm" {
  features {}
  subscription_id = var.azure_subscription_id
}

provider "google" {
  project = var.gcp_project_id
  region  = "us-central1"
}

# AWS EKS Cluster
module "aws_eks" {
  source = "./modules/eks"
  providers = {
    aws = aws.us_east
  }

  cluster_name = "multi-cloud-eks"
  cluster_version = "1.27"
  vpc_id = module.aws_vpc.vpc_id
  subnet_ids = module.aws_vpc.private_subnet_ids
}

# Azure AKS Cluster
module "azure_aks" {
  source = "./modules/aks"
  providers = {
    azurerm = azurerm
  }

  cluster_name = "multi-cloud-aks"
  kubernetes_version = "1.27"
  vnet_id = module.azure_vnet.vnet_id
  subnet_ids = module.azure_vnet.subnet_ids
}

# GCP GKE Cluster
module "gcp_gke" {
  source = "./modules/gke"
  providers = {
    google = google
  }

  cluster_name = "multi-cloud-gke"
  kubernetes_version = "1.27"
  network = module.gcp_vpc.network_id
  subnetwork = module.gcp_vpc.subnet_id
}

# Multi-cloud networking
module "multi_cloud_network" {
  source = "./modules/multi-cloud-network"

  aws_vpc_id = module.aws_vpc.vpc_id
  azure_vnet_id = module.azure_vnet.vnet_id
  gcp_network_id = module.gcp_vpc.network_id

  aws_subnet_ids = module.aws_vpc.private_subnet_ids
  azure_subnet_ids = module.azure_vnet.subnet_ids
  gcp_subnet_id = module.gcp_vpc.subnet_id
}

# Global load balancer
module "global_load_balancer" {
  source = "./modules/global-lb"

  aws_endpoints = module.aws_eks.endpoint
  azure_endpoints = module.azure_aks.endpoint
  gcp_endpoints = module.gcp_gke.endpoint
}

Kubernetes Multi-Cloud Abstraction

# Multi-cloud Kubernetes deployment
apiVersion: apps/v1
kind: Deployment
metadata:
  name: multi-cloud-app
  labels:
    app: multi-cloud-app
    cloud: ${CLOUD_PROVIDER}
spec:
  replicas: 3
  selector:
    matchLabels:
      app: multi-cloud-app
  template:
    metadata:
      labels:
        app: multi-cloud-app
        cloud: ${CLOUD_PROVIDER}
    spec:
      # Cloud-specific node selector
      nodeSelector:
        cloud: ${CLOUD_PROVIDER}
      
      # Cloud-specific tolerations
      tolerations:
        - key: "cloud"
          operator: "Equal"
          value: "${CLOUD_PROVIDER}"
          effect: "NoSchedule"
      
      containers:
        - name: app
          image: multi-cloud-app:latest
          ports:
            - containerPort: 8080
          
          # Cloud-specific resources
          resources:
            requests:
              cpu: "500m"
              memory: "512Mi"
              # AWS-specific
              # aws.amazon.com/gpu: "1"
              # Azure-specific
              # azure.com/gpu: "1"
              # GCP-specific
              # cloud.google.com/gpu: "1"
          
          # Cloud-specific environment variables
          env:
            - name: CLOUD_PROVIDER
              value: ${CLOUD_PROVIDER}
            - name: CLOUD_REGION
              value: ${CLOUD_REGION}
            - name: CLOUD_ZONE
              value: ${CLOUD_ZONE}
          
          # Cloud-specific volume mounts
          volumeMounts:
            - name: cloud-storage
              mountPath: /data
      
      volumes:
        - name: cloud-storage
          # Cloud-specific volume type
          # AWS EBS
          # awsElasticBlockStore:
          #   volumeID: vol-12345
          #   fsType: ext4
          
          # Azure Disk
          # azureDisk:
          #   diskName: my-disk
          #   diskURI: /subscriptions/xxx/resourceGroups/xxx/providers/Microsoft.Compute/disks/my-disk
          
          # GCP Persistent Disk
          # gcePersistentDisk:
          #   pdName: my-disk
          #   fsType: ext4
          
          emptyDir: {}

---
# Cloud-agnostic service
apiVersion: v1
kind: Service
metadata:
  name: multi-cloud-service
  annotations:
    # AWS ALB
    # service.beta.kubernetes.io/aws-load-balancer-type: "nlb"
    
    # Azure Load Balancer
    # service.beta.kubernetes.io/azure-load-balancer-internal: "true"
    
    # GCP Load Balancer
    # cloud.google.com/neg: '{"ingress": true}'
spec:
  selector:
    app: multi-cloud-app
  ports:
    - port: 80
      targetPort: 8080
  type: LoadBalancer

Cloud Abstraction Layer

# Cloud abstraction layer
from abc import ABC, abstractmethod
from typing import Dict, Any, Optional, List
from dataclasses import dataclass
from enum import Enum

class CloudProvider(Enum):
    AWS = "aws"
    AZURE = "azure"
    GCP = "gcp"

@dataclass
class CloudConfig:
    provider: CloudProvider
    region: str
    credentials: Dict[str, str]

class CloudProviderInterface(ABC):
    """Abstract interface for cloud providers"""

    @abstractmethod
    def create_vm(self, name: str, size: str, image: str) -> str:
        pass

    @abstractmethod
    def create_database(self, name: str, engine: str, size: str) -> str:
        pass

    @abstractmethod
    def create_storage(self, name: str) -> str:
        pass

    @abstractmethod
    def deploy_container(self, name: str, image: str, replicas: int) -> str:
        pass

    @abstractmethod
    def get_metrics(self, resource_id: str) -> Dict[str, Any]:
        pass

class AWSProvider(CloudProviderInterface):
    """AWS implementation"""

    def __init__(self, config: CloudConfig):
        self.config = config
        self.ec2 = boto3.client('ec2', region_name=config.region)
        self.rds = boto3.client('rds', region_name=config.region)
        self.s3 = boto3.client('s3', region_name=config.region)
        self.ecs = boto3.client('ecs', region_name=config.region)

    def create_vm(self, name: str, size: str, image: str) -> str:
        response = self.ec2.run_instances(
            ImageId=image,
            InstanceType=size,
            MinCount=1,
            MaxCount=1,
            TagSpecifications=[{
                'ResourceType': 'instance',
                'Tags': [{'Key': 'Name', 'Value': name}]
            }]
        )
        return response['Instances'][0]['InstanceId']

    def create_database(self, name: str, engine: str, size: str) -> str:
        response = self.rds.create_db_instance(
            DBInstanceIdentifier=name,
            DBInstanceClass=size,
            Engine=engine,
            MasterUsername='admin',
            MasterUserPassword='password'
        )
        return response['DBInstance']['DBInstanceIdentifier']

    def create_storage(self, name: str) -> str:
        self.s3.create_bucket(Bucket=name)
        return name

    def deploy_container(self, name: str, image: str, replicas: int) -> str:
        response = self.ecs.create_service(
            cluster='default',
            serviceName=name,
            taskDefinition=image,
            desiredCount=replicas
        )
        return response['service']['serviceName']

    def get_metrics(self, resource_id: str) -> Dict[str, Any]:
        cloudwatch = boto3.client('cloudwatch', region_name=self.config.region)
        response = cloudwatch.get_metric_statistics(
            Namespace='AWS/EC2',
            MetricName='CPUUtilization',
            Dimensions=[{'Name': 'InstanceId', 'Value': resource_id}],
            StartTime=datetime.utcnow() - timedelta(hours=1),
            EndTime=datetime.utcnow(),
            Period=300,
            Statistics=['Average']
        )
        return response

class AzureProvider(CloudProviderInterface):
    """Azure implementation"""

    def __init__(self, config: CloudConfig):
        self.config = config
        # Azure SDK initialization

    def create_vm(self, name: str, size: str, image: str) -> str:
        # Azure VM creation
        return f"azure-vm-{name}"

    def create_database(self, name: str, engine: str, size: str) -> str:
        # Azure SQL Database creation
        return f"azure-db-{name}"

    def create_storage(self, name: str) -> str:
        # Azure Blob Storage creation
        return f"azure-storage-{name}"

    def deploy_container(self, name: str, image: str, replicas: int) -> str:
        # Azure Container Instances
        return f"azure-container-{name}"

    def get_metrics(self, resource_id: str) -> Dict[str, Any]:
        # Azure Monitor metrics
        return {}

class GCPProvider(CloudProviderInterface):
    """GCP implementation"""

    def __init__(self, config: CloudConfig):
        self.config = config
        # GCP SDK initialization

    def create_vm(self, name: str, size: str, image: str) -> str:
        # GCP Compute Engine
        return f"gcp-vm-{name}"

    def create_database(self, name: str, engine: str, size: str) -> str:
        # GCP Cloud SQL
        return f"gcp-db-{name}"

    def create_storage(self, name: str) -> str:
        # GCP Cloud Storage
        return f"gcp-storage-{name}"

    def deploy_container(self, name: str, image: str, replicas: int) -> str:
        # GCP GKE
        return f"gcp-container-{name}"

    def get_metrics(self, resource_id: str) -> Dict[str, Any]:
        # GCP Cloud Monitoring
        return {}

class MultiCloudManager:
    """Multi-cloud management"""

    def __init__(self):
        self.providers: Dict[CloudProvider, CloudProviderInterface] = {}

    def register_provider(self, provider: CloudProvider, interface: CloudProviderInterface):
        """Register cloud provider"""
        self.providers[provider] = interface

    def deploy_to_cloud(self, provider: CloudProvider, 
                       resource_type: str, **kwargs) -> str:
        """Deploy resource to specific cloud"""
        interface = self.providers.get(provider)
        if not interface:
            raise ValueError(f"Provider {provider} not registered")

        if resource_type == 'vm':
            return interface.create_vm(**kwargs)
        elif resource_type == 'database':
            return interface.create_database(**kwargs)
        elif resource_type == 'storage':
            return interface.create_storage(**kwargs)
        elif resource_type == 'container':
            return interface.deploy_container(**kwargs)
        else:
            raise ValueError(f"Unknown resource type: {resource_type}")

    def deploy_multi_cloud(self, resource_type: str, 
                          distribution: Dict[CloudProvider, int], **kwargs) -> Dict[str, str]:
        """Deploy across multiple clouds"""
        results = {}

        for provider, count in distribution.items():
            for i in range(count):
                resource_id = self.deploy_to_cloud(
                    provider,
                    resource_type,
                    **{**kwargs, 'name': f"{kwargs['name']}-{provider.value}-{i}"}
                )
                results[f"{provider.value}-{i}"] = resource_id

        return results

Cost Optimization Across Clouds

# Multi-cloud cost optimization
import boto3
from typing import Dict, Any, List
from dataclasses import dataclass
from datetime import datetime, timedelta

@dataclass
class CloudCostReport:
    provider: str
    service: str
    cost: float
    period: str
    trend: float

class MultiCloudCostOptimizer:
    """Multi-cloud cost optimization"""

    def __init__(self):
        self.providers = {
            'aws': self._get_aws_cost,
            'azure': self._get_azure_cost,
            'gcp': self._get_gcp_cost
        }

    def get_total_cost(self, days: int = 30) -> Dict[str, Any]:
        """Get total cost across all clouds"""
        costs = {}
        total = 0

        for provider, getter in self.providers.items():
            cost = getter(days)
            costs[provider] = cost
            total += cost

        return {
            'costs': costs,
            'total': total,
            'period': f'last_{days}_days'
        }

    def _get_aws_cost(self, days: int) -> float:
        """Get AWS cost"""
        ce = boto3.client('ce')
        end_date = datetime.utcnow().strftime('%Y-%m-%d')
        start_date = (datetime.utcnow() - timedelta(days=days)).strftime('%Y-%m-%d')

        response = ce.get_cost_and_usage(
            TimePeriod={'Start': start_date, 'End': end_date},
            Granularity='MONTHLY',
            Metrics=['UnblendedCost']
        )

        total = sum(
            float(result['Total']['UnblendedCost']['Amount'])
            for result in response['ResultsByTime']
        )

        return total

    def _get_azure_cost(self, days: int) -> float:
        """Get Azure cost"""
        # Azure Cost Management API
        return 0.0

    def _get_gcp_cost(self, days: int) -> float:
        """Get GCP cost"""
        # GCP Billing API
        return 0.0

    def optimize_cloud_distribution(self) -> Dict[str, Any]:
        """Optimize workload distribution across clouds"""
        costs = self.get_total_cost()

        # Analyze cost efficiency
        efficiencies = {}
        for provider, cost in costs['costs'].items():
            # Assume same workload across all clouds
            efficiencies[provider] = 1 / cost if cost > 0 else 0

        # Normalize to percentages
        total_efficiency = sum(efficiencies.values())
        optimal_distribution = {
            provider: efficiency / total_efficiency
            for provider, efficiency in efficiencies.items()
        }

        return {
            'current_distribution': self._get_current_distribution(),
            'optimal_distribution': optimal_distribution,
            'potential_savings': self._calculate_savings(optimal_distribution)
        }

    def _get_current_distribution(self) -> Dict[str, float]:
        """Get current cloud distribution"""
        return {'aws': 0.4, 'azure': 0.35, 'gcp': 0.25}

    def _calculate_savings(self, optimal: Dict[str, float]) -> float:
        """Calculate potential savings"""
        current = self._get_current_distribution()
        costs = self.get_total_cost()

        current_cost = sum(
            costs['costs'].get(provider, 0) * share
            for provider, share in current.items()
        )

        optimal_cost = sum(
            costs['costs'].get(provider, 0) * share
            for provider, share in optimal.items()
        )

        return current_cost - optimal_cost

⚠️Multi-Cloud Considerations

Multi-cloud adds complexity. Use it for specific needs like compliance, performance, or vendor lock-in avoidance. Start with a primary cloud and add others strategically.

Summary

Strategy	Purpose	Implementation
Abstraction	Cloud-agnostic	Terraform, Kubernetes
Portability	Move workloads	Containers, microservices
Cost Optimization	Reduce spend	Multi-cloud analytics
Governance	Compliance	Policy-as-code
Disaster Recovery	High availability	Multi-region, multi-cloud