AI Model Wiki

    Skip to main content
    22 min
    Lesson 3 of 4

    Governance and Operations

    From the course Multi-Cloud AI Strategy

    Operate and govern multi-cloud AI infrastructure with unified tooling and policies.

    Unified Governance

    Infrastructure as Code

    # Terraform for multi-cloud AI deployment
    # AWS Bedrock
    resource "aws_bedrock_provisioned_model_throughput" "main" {
      model_id = "anthropic.claude-3-5-sonnet"
      model_units = 1
    }
    
    # Azure OpenAI
    resource "azurerm_cognitive_deployment" "main" {
      name = "gpt-4o-deployment"
      cognitive_account_id = azurerm_cognitive_account.main.id
      model {
        format = "OpenAI"
        name = "gpt-4o"
        version = "2024-08-06"
      }
    }
    
    # GCP Vertex AI
    resource "google_vertex_ai_endpoint" "main" {
      name = "gemini-endpoint"
      region = "us-central1"
    }
    

    Unified Monitoring Dashboard

    # Aggregate metrics from all cloud providers
    class UnifiedAIDashboard:
        def get_all_metrics(self):
            return {
                "aws": self.get_bedrock_metrics(),
                "azure": self.get_azure_openai_metrics(),
                "gcp": self.get_vertex_metrics()
            }
        
        def get_total_daily_cost(self):
            metrics = self.get_all_metrics()
            return sum(m["daily_cost"] for m in metrics.values())
        
        def get_aggregate_sla_status(self):
            # Return worst SLA across all providers
            metrics = self.get_all_metrics()
            return min(m["availability"] for m in metrics.values())
    

    Multi-Cloud Best Practices

    Decision Framework

    FactorAWSAzureGCP
    LLM Diversity★★★★★★★★★★★
    Microsoft Integration★★★★★★★
    Google AI Models★★★★★★★
    MLOps Maturity★★★★★★★★★★★
    Cost Efficiency★★★★★★★★★★★

    Governance Policies

    • Unified tagging standards across all clouds
    • Single identity provider (Azure AD or Okta)
    • Centralized audit logging to SIEM
    • Consistent data classification and handling
    • Multi-cloud incident response runbooks

    Key Takeaways

    • •Terraform enables consistent multi-cloud AI infrastructure provisioning
    • •Unified monitoring dashboard aggregates metrics and costs from all providers
    • •Provider selection matrix guides workload placement decisions
    • •Single identity provider and unified tagging simplify multi-cloud governance
    0 of 4 lessons marked complete

    Progress is kept in this browser only.