Internet
    │
    ▼
┌─────────────────────────────────────────────────────────────┐
│                    CloudFront CDN                           │
│                 (Global Distribution)                       │
└─────────────────────────────────────────────────────────────┘
    │
    ▼
┌─────────────────────────────────────────────────────────────┐
│                Application Load Balancer                    │
│              (Multi-AZ, Auto Scaling)                      │
└─────────────────────────────────────────────────────────────┘
    │
    ▼
┌─────────────────────────────────────────────────────────────┐
│                    ECS Fargate Cluster                     │
│   ┌─────────────┐  ┌─────────────┐  ┌─────────────┐       │
│   │   Django    │  │   Celery    │  │   Celery    │       │
│   │  Web Tasks  │  │   Workers   │  │    Beat     │       │
│   └─────────────┘  └─────────────┘  └─────────────┘       │
└─────────────────────────────────────────────────────────────┘
    │                     │                     │
    ▼                     ▼                     ▼
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│   RDS PostgreSQL│ │ ElastiCache Redis│ │      S3         │
│   (Multi-AZ)    │ │   (Clustered)    │ │  (Media/Static) │
└─────────────────┘ └─────────────────┘ └─────────────────┘

{
  "cluster_name": "jobhive-production",
  "launch_type": "FARGATE",
  "capacity_providers": ["FARGATE", "FARGATE_SPOT"],
  "default_capacity_provider_strategy": [
    {
      "capacity_provider": "FARGATE",
      "weight": 4,
      "base": 2
    },
    {
      "capacity_provider": "FARGATE_SPOT",
      "weight": 1
    }
  ]
}

service_name: jobhive-web
task_definition: jobhive-web-task:latest
launch_type: FARGATE
desired_count: 3
platform_version: LATEST

network_configuration:
  subnets:
    - subnet-1a2b3c4d
    - subnet-5e6f7g8h
  security_groups:
    - sg-web-application
  assign_public_ip: DISABLED

load_balancers:
  - target_group_arn: arn:aws:elasticloadbalancing:us-east-1:123456789:targetgroup/jobhive-web
    container_name: django-web
    container_port: 8000

auto_scaling:
  min_capacity: 2
  max_capacity: 10
  target_cpu_utilization: 70
  target_memory_utilization: 80

service_name: jobhive-celery-workers
task_definition: jobhive-celery-worker-task:latest
launch_type: FARGATE
desired_count: 2

auto_scaling:
  min_capacity: 1
  max_capacity: 8
  target_cpu_utilization: 75
  custom_metrics:
    - metric_name: celery_queue_length
      target_value: 10

service_name: jobhive-celery-beat
task_definition: jobhive-celery-beat-task:latest
launch_type: FARGATE
desired_count: 1  # Single instance for scheduled tasks

deployment_configuration:
  maximum_percent: 100
  minimum_healthy_percent: 0  # Allow complete replacement

{
  "family": "jobhive-web-task",
  "network_mode": "awsvpc",
  "requires_compatibilities": ["FARGATE"],
  "cpu": "512",
  "memory": "1024",
  "execution_role_arn": "arn:aws:iam::123456789:role/ecsTaskExecutionRole",
  "task_role_arn": "arn:aws:iam::123456789:role/jobhiveTaskRole",
  "container_definitions": [
    {
      "name": "django-web",
      "image": "123456789.dkr.ecr.us-east-1.amazonaws.com/jobhive/backend:latest",
      "port_mappings": [
        {
          "container_port": 8000,
          "protocol": "tcp"
        }
      ],
      "environment": [
        {"name": "DJANGO_SETTINGS_MODULE", "value": "config.settings.production"},
        {"name": "AWS_DEFAULT_REGION", "value": "us-east-1"}
      ],
      "secrets": [
        {"name": "DATABASE_URL", "value_from": "arn:aws:secretsmanager:us-east-1:123456789:secret:jobhive/database"},
        {"name": "REDIS_URL", "value_from": "arn:aws:secretsmanager:us-east-1:123456789:secret:jobhive/redis"},
        {"name": "DJANGO_SECRET_KEY", "value_from": "arn:aws:secretsmanager:us-east-1:123456789:secret:jobhive/django"}
      ],
      "log_configuration": {
        "log_driver": "awslogs",
        "options": {
          "awslogs-group": "/ecs/jobhive-web",
          "awslogs-region": "us-east-1",
          "awslogs-stream-prefix": "ecs"
        }
      },
      "health_check": {
        "command": ["CMD-SHELL", "curl -f http://localhost:8000/health/ || exit 1"],
        "interval": 30,
        "timeout": 5,
        "retries": 3,
        "start_period": 60
      }
    }
  ]
}

engine: postgres
engine_version: "15.4"
instance_class: db.r6g.xlarge
allocated_storage: 1000
storage_type: gp3
storage_encrypted: true
kms_key_id: alias/rds-encryption-key

multi_az: true
availability_zones:
  - us-east-1a
  - us-east-1b

backup_retention_period: 30
backup_window: "03:00-04:00"
maintenance_window: "sun:04:00-sun:05:00"

monitoring_interval: 60
performance_insights_enabled: true
performance_insights_retention_period: 731

parameter_group: jobhive-postgres15-params
option_group: default:postgres-15

vpc_security_group_ids:
  - sg-rds-database

db_subnet_group: jobhive-db-subnet-group

-- PostgreSQL parameter group settings
shared_preload_libraries = 'pg_stat_statements'
max_connections = 200
shared_buffers = '4GB'
effective_cache_size = '12GB'
work_mem = '64MB'
maintenance_work_mem = '512MB'
random_page_cost = 1.1
effective_io_concurrency = 200
wal_buffers = '16MB'
checkpoint_completion_target = 0.9
max_wal_size = '4GB'
min_wal_size = '1GB'

-- Performance optimization settings
log_statement = 'mod'
log_min_duration_statement = 1000
track_activity_query_size = 2048
pg_stat_statements.track = all

read_replica_1:
  instance_class: db.r6g.large
  availability_zone: us-east-1c
  publicly_accessible: false
  auto_minor_version_upgrade: true
  
read_replica_2:
  instance_class: db.r6g.large  
  availability_zone: us-east-1d
  publicly_accessible: false
  auto_minor_version_upgrade: true

cache_cluster_id: jobhive-redis-cluster
engine: redis
engine_version: "7.0"
node_type: cache.r7g.large
num_cache_nodes: 3

replication_group_id: jobhive-redis-replication
description: "JobHive Redis cluster for caching and sessions"
port: 6379

parameter_group_name: jobhive-redis7-params
subnet_group_name: jobhive-cache-subnet-group
security_group_ids:
  - sg-redis-cache

at_rest_encryption_enabled: true
transit_encryption_enabled: true
auth_token: "stored-in-secrets-manager"

automatic_failover_enabled: true
multi_az_enabled: true

backup_retention_limit: 7
backup_window: "02:00-03:00"
maintenance_window: "sun:01:00-sun:02:00"

notification_topic_arn: "arn:aws:sns:us-east-1:123456789:jobhive-cache-notifications"

# Redis configuration parameters
maxmemory-policy allkeys-lru
maxmemory 6gb
save 900 1
save 300 10  
save 60 10000
timeout 300
tcp-keepalive 300
databases 16

# Primary media bucket
media_bucket:
  name: jobhive-media-production
  region: us-east-1
  versioning: Enabled
  encryption:
    sse_algorithm: AES256
  lifecycle_policy:
    - id: delete_incomplete_multipart_uploads
      status: Enabled
      abort_incomplete_multipart_upload:
        days_after_initiation: 7
    - id: transition_to_ia
      status: Enabled
      transition:
        days: 30
        storage_class: STANDARD_IA
    - id: transition_to_glacier
      status: Enabled
      transition:
        days: 90
        storage_class: GLACIER

# Static assets bucket  
static_bucket:
  name: jobhive-static-production
  region: us-east-1
  public_read_access: true
  website_configuration:
    index_document: index.html
    error_document: error.html
  cors_configuration:
    - allowed_origins: ["https://jobhive.com", "https://app.jobhive.com"]
      allowed_methods: ["GET", "HEAD"]
      allowed_headers: ["*"]
      max_age_seconds: 3600

# Backup bucket
backup_bucket:
  name: jobhive-backups-production
  region: us-west-2  # Different region for disaster recovery
  versioning: Enabled
  encryption:
    sse_algorithm: aws:kms
    kms_master_key_id: alias/jobhive-backup-key
  lifecycle_policy:
    - id: delete_old_backups
      status: Enabled
      expiration:
        days: 2555  # 7 years retention

distribution:
  enabled: true
  price_class: PriceClass_All
  http_version: http2
  is_ipv6_enabled: true
  
  origins:
    - id: s3-static-origin
      domain_name: jobhive-static-production.s3.amazonaws.com
      s3_origin_config:
        origin_access_identity: E1234567890ABC
    - id: alb-api-origin
      domain_name: jobhive-alb-1234567890.us-east-1.elb.amazonaws.com
      custom_origin_config:
        http_port: 443
        origin_protocol_policy: https-only
        origin_ssl_protocols: [TLSv1.2]

  default_cache_behavior:
    target_origin_id: alb-api-origin
    viewer_protocol_policy: redirect-to-https
    allowed_methods: [GET, HEAD, OPTIONS, PUT, POST, PATCH, DELETE]
    cached_methods: [GET, HEAD, OPTIONS]
    compress: true
    
    forwarded_values:
      query_string: true
      headers: [Authorization, Content-Type]
      cookies:
        forward: whitelist
        whitelisted_names: [sessionid, csrftoken]

  cache_behaviors:
    - path_pattern: "/static/*"
      target_origin_id: s3-static-origin
      viewer_protocol_policy: redirect-to-https
      allowed_methods: [GET, HEAD]
      cached_methods: [GET, HEAD]
      compress: true
      min_ttl: 86400
      default_ttl: 31536000
      max_ttl: 31536000

  restrictions:
    geo_restriction:
      restriction_type: none

  viewer_certificate:
    acm_certificate_arn: arn:aws:acm:us-east-1:123456789:certificate/12345678-1234-1234-1234-123456789012
    ssl_support_method: sni-only
    minimum_protocol_version: TLSv1.2_2021

load_balancer:
  name: jobhive-alb
  scheme: internet-facing
  type: application
  ip_address_type: ipv4
  
  subnets:
    - subnet-1a2b3c4d  # Public subnet AZ-a
    - subnet-5e6f7g8h  # Public subnet AZ-b
    - subnet-9i0j1k2l  # Public subnet AZ-c
    
  security_groups:
    - sg-alb-external

  listeners:
    - port: 443
      protocol: HTTPS
      ssl_policy: ELBSecurityPolicy-TLS-1-2-2017-01
      certificate_arn: arn:aws:acm:us-east-1:123456789:certificate/12345678-1234-1234-1234-123456789012
      default_actions:
        - type: forward
          target_group_arn: arn:aws:elasticloadbalancing:us-east-1:123456789:targetgroup/jobhive-web
    
    - port: 80
      protocol: HTTP
      default_actions:
        - type: redirect
          redirect:
            protocol: HTTPS
            port: 443
            status_code: HTTP_301

target_groups:
  - name: jobhive-web
    port: 8000
    protocol: HTTP
    target_type: ip
    vpc_id: vpc-12345678
    
    health_check:
      enabled: true
      healthy_threshold_count: 2
      unhealthy_threshold_count: 3
      timeout: 5
      interval: 30
      path: /health/
      matcher: 200
      protocol: HTTP
      port: traffic-port

vpc:
  cidr_block: 10.0.0.0/16
  enable_dns_hostnames: true
  enable_dns_support: true
  
  # Public subnets for ALB and NAT Gateway
  public_subnets:
    - cidr: 10.0.1.0/24
      availability_zone: us-east-1a
    - cidr: 10.0.2.0/24
      availability_zone: us-east-1b
    - cidr: 10.0.3.0/24
      availability_zone: us-east-1c
  
  # Private subnets for ECS tasks
  private_subnets:
    - cidr: 10.0.10.0/24
      availability_zone: us-east-1a
    - cidr: 10.0.11.0/24
      availability_zone: us-east-1b
    - cidr: 10.0.12.0/24
      availability_zone: us-east-1c
  
  # Database subnets
  database_subnets:
    - cidr: 10.0.20.0/24
      availability_zone: us-east-1a
    - cidr: 10.0.21.0/24
      availability_zone: us-east-1b
    - cidr: 10.0.22.0/24
      availability_zone: us-east-1c

  # Internet Gateway
  internet_gateway:
    name: jobhive-igw
  
  # NAT Gateways for private subnet internet access
  nat_gateways:
    - name: jobhive-nat-1a
      subnet: public-subnet-1a
      allocation_id: eipalloc-12345678
    - name: jobhive-nat-1b
      subnet: public-subnet-1b
      allocation_id: eipalloc-87654321

# ALB Security Group
sg_alb_external:
  name: jobhive-alb-external
  description: Security group for ALB
  vpc_id: vpc-12345678
  ingress_rules:
    - from_port: 80
      to_port: 80
      protocol: tcp
      cidr_blocks: [0.0.0.0/0]
    - from_port: 443
      to_port: 443
      protocol: tcp
      cidr_blocks: [0.0.0.0/0]
  egress_rules:
    - from_port: 0
      to_port: 65535
      protocol: tcp
      cidr_blocks: [10.0.0.0/16]

# ECS Tasks Security Group
sg_web_application:
  name: jobhive-web-application
  description: Security group for ECS web tasks
  vpc_id: vpc-12345678
  ingress_rules:
    - from_port: 8000
      to_port: 8000
      protocol: tcp
      source_security_group_id: sg-alb-external
  egress_rules:
    - from_port: 0
      to_port: 65535
      protocol: tcp
      cidr_blocks: [0.0.0.0/0]

# RDS Security Group
sg_rds_database:
  name: jobhive-rds-database
  description: Security group for RDS PostgreSQL
  vpc_id: vpc-12345678
  ingress_rules:
    - from_port: 5432
      to_port: 5432
      protocol: tcp
      source_security_group_id: sg-web-application
  egress_rules: []

# Redis Security Group
sg_redis_cache:
  name: jobhive-redis-cache
  description: Security group for ElastiCache Redis
  vpc_id: vpc-12345678
  ingress_rules:
    - from_port: 6379
      to_port: 6379
      protocol: tcp
      source_security_group_id: sg-web-application
  egress_rules: []

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "ecr:GetAuthorizationToken",
        "ecr:BatchCheckLayerAvailability",
        "ecr:GetDownloadUrlForLayer",
        "ecr:BatchGetImage",
        "logs:CreateLogStream",
        "logs:PutLogEvents",
        "secretsmanager:GetSecretValue"
      ],
      "Resource": "*"
    }
  ]
}

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "s3:GetObject",
        "s3:PutObject",
        "s3:DeleteObject"
      ],
      "Resource": [
        "arn:aws:s3:::jobhive-media-production/*",
        "arn:aws:s3:::jobhive-static-production/*"
      ]
    },
    {
      "Effect": "Allow",
      "Action": [
        "s3:ListBucket"
      ],
      "Resource": [
        "arn:aws:s3:::jobhive-media-production",
        "arn:aws:s3:::jobhive-static-production"
      ]
    },
    {
      "Effect": "Allow",
      "Action": [
        "secretsmanager:GetSecretValue"
      ],
      "Resource": [
        "arn:aws:secretsmanager:us-east-1:123456789:secret:jobhive/*"
      ]
    },
    {
      "Effect": "Allow",
      "Action": [
        "ssm:GetParameter",
        "ssm:GetParameters"
      ],
      "Resource": [
        "arn:aws:ssm:us-east-1:123456789:parameter/jobhive/*"
      ]
    }
  ]
}

secrets:
  database_credentials:
    name: jobhive/database
    description: PostgreSQL database credentials
    secret_string:
      engine: postgres
      host: jobhive-db.cluster-xyz.us-east-1.rds.amazonaws.com
      username: jobhive_app
      password: "auto-generated-secure-password"
      dbname: jobhive_production
      port: 5432

  redis_credentials:
    name: jobhive/redis
    description: Redis cache credentials
    secret_string:
      host: jobhive-redis-cluster.abc123.cache.amazonaws.com
      port: 6379
      auth_token: "auto-generated-auth-token"

  django_secrets:
    name: jobhive/django
    description: Django application secrets
    secret_string:
      secret_key: "auto-generated-django-secret-key"
      jwt_secret: "auto-generated-jwt-secret"

  stripe_keys:
    name: jobhive/stripe
    description: Stripe payment processing keys
    secret_string:
      publishable_key: pk_live_...
      secret_key: sk_live_...
      webhook_secret: whsec_...

  external_services:
    name: jobhive/external
    description: External service API keys
    secret_string:
      openai_api_key: sk-...
      datadog_api_key: ...
      aws_transcribe_region: us-east-1

log_groups:
  - name: /ecs/jobhive-web
    retention_in_days: 30
    kms_key_id: arn:aws:kms:us-east-1:123456789:key/12345678-1234-1234-1234-123456789012
  
  - name: /ecs/jobhive-celery-worker
    retention_in_days: 30
    kms_key_id: arn:aws:kms:us-east-1:123456789:key/12345678-1234-1234-1234-123456789012
  
  - name: /ecs/jobhive-celery-beat
    retention_in_days: 7
    kms_key_id: arn:aws:kms:us-east-1:123456789:key/12345678-1234-1234-1234-123456789012

  - name: /aws/rds/instance/jobhive-db/postgresql
    retention_in_days: 30

  - name: /aws/elasticache/jobhive-redis-cluster
    retention_in_days: 14

alarms:
  # Application health alarms
  - name: jobhive-web-high-cpu
    description: Web service CPU utilization is high
    metric_name: CPUUtilization
    namespace: AWS/ECS
    statistic: Average
    period: 300
    evaluation_periods: 2
    threshold: 80
    comparison_operator: GreaterThanThreshold
    alarm_actions:
      - arn:aws:sns:us-east-1:123456789:jobhive-alerts

  - name: jobhive-web-high-memory
    description: Web service memory utilization is high
    metric_name: MemoryUtilization
    namespace: AWS/ECS
    statistic: Average
    period: 300
    evaluation_periods: 2
    threshold: 85
    comparison_operator: GreaterThanThreshold

  # Database alarms
  - name: jobhive-db-high-cpu
    description: Database CPU utilization is high
    metric_name: CPUUtilization
    namespace: AWS/RDS
    statistic: Average
    period: 300
    evaluation_periods: 3
    threshold: 75
    comparison_operator: GreaterThanThreshold

  - name: jobhive-db-low-freeable-memory
    description: Database freeable memory is low
    metric_name: FreeableMemory
    namespace: AWS/RDS
    statistic: Average
    period: 300
    evaluation_periods: 2
    threshold: 1000000000  # 1GB in bytes
    comparison_operator: LessThanThreshold

  # Cache alarms
  - name: jobhive-redis-high-cpu
    description: Redis CPU utilization is high
    metric_name: CPUUtilization
    namespace: AWS/ElastiCache
    statistic: Average
    period: 300
    evaluation_periods: 2
    threshold: 80
    comparison_operator: GreaterThanThreshold

datadog_agent:
  deployment_type: sidecar
  api_key_secret: arn:aws:secretsmanager:us-east-1:123456789:secret:jobhive/datadog
  
  configuration:
    logs_enabled: true
    process_agent_enabled: true
    system_probe_enabled: true
    
    integrations:
      - postgres:
          host: "%%host%%"
          port: 5432
          username: datadog
          password: "%%env_DD_POSTGRES_PASSWORD%%"
          tags:
            - environment:production
            - service:jobhive-db
      
      - redisdb:
          host: "%%host%%"
          port: 6379
          password: "%%env_DD_REDIS_PASSWORD%%"
          tags:
            - environment:production
            - service:jobhive-cache

    tags:
      - environment:production
      - service:jobhive
      - version:1.0.0

auto_scaling_policies:
  # Web service scaling
  - service_name: jobhive-web
    policies:
      - name: cpu-scaling
        policy_type: TargetTrackingScaling
        target_tracking_scaling_policy:
          target_value: 70.0
          predefined_metric_specification:
            predefined_metric_type: ECSServiceAverageCPUUtilization
          scale_out_cooldown: 300
          scale_in_cooldown: 300
      
      - name: memory-scaling
        policy_type: TargetTrackingScaling
        target_tracking_scaling_policy:
          target_value: 80.0
          predefined_metric_specification:
            predefined_metric_type: ECSServiceAverageMemoryUtilization
          scale_out_cooldown: 300
          scale_in_cooldown: 600
  
  # Celery worker scaling
  - service_name: jobhive-celery-workers
    policies:
      - name: queue-based-scaling
        policy_type: StepScaling
        step_scaling_policy:
          adjustment_type: ChangeInCapacity
          cooldown: 300
          step_adjustments:
            - lower_bound: 0
              upper_bound: 5
              scaling_adjustment: 1
            - lower_bound: 5
              upper_bound: 15
              scaling_adjustment: 2
            - lower_bound: 15
              scaling_adjustment: 3

target_group_scaling:
  target_group: jobhive-web
  health_check_grace_period: 300
  
  scaling_policies:
    - metric_type: request_count_per_target
      target_value: 1000
      scale_out_cooldown: 300
      scale_in_cooldown: 300

rds_backup:
  automated_backups:
    backup_retention_period: 30
    backup_window: "03:00-04:00"
    copy_tags_to_snapshot: true
    delete_automated_backups: true
    
  manual_snapshots:
    - schedule: daily
      retention: 7_days
      encryption: true
    - schedule: weekly  
      retention: 4_weeks
      encryption: true
    - schedule: monthly
      retention: 12_months
      encryption: true
      cross_region_copy:
        destination_region: us-west-2
        kms_key_id: alias/jobhive-backup-key-west

point_in_time_recovery:
  enabled: true
  earliest_restorable_time: current_time - 30_days

application_backup:
  s3_media_backup:
    source_bucket: jobhive-media-production
    destination_bucket: jobhive-backups-production
    schedule: daily
    retention: 90_days
    
    cross_region_replication:
      destination_bucket: jobhive-backups-west
      destination_region: us-west-2
      storage_class: STANDARD_IA

  configuration_backup:
    ecs_task_definitions: daily
    cloudformation_templates: on_change
    secrets_manager: weekly
    parameter_store: weekly

disaster_recovery:
  rto: 4_hours  # Recovery Time Objective
  rpo: 15_minutes  # Recovery Point Objective

  primary_region: us-east-1
  secondary_region: us-west-2

  failover_procedures:
    1_database_failover:
      - create_rds_instance_from_snapshot
      - update_dns_records
      - validate_data_integrity
      
    2_application_failover:
      - deploy_ecs_services_in_secondary_region
      - update_load_balancer_targets
      - update_cloudfront_origins
      
    3_data_synchronization:
      - restore_s3_media_from_backup
      - validate_application_functionality
      - update_monitoring_dashboards

  testing_schedule:
    frequency: quarterly
    scope: full_failover_test
    documentation: required

cost_optimization:
  compute:
    - use_spot_instances: 30%_of_celery_workers
    - scheduled_scaling: scale_down_non_business_hours
    - instance_right_sizing: quarterly_review
    
  storage:
    - s3_intelligent_tiering: enabled
    - lifecycle_policies: automated_transition
    - compress_backups: enabled
    - delete_unused_snapshots: automated
    
  data_transfer:
    - cloudfront_caching: aggressive_static_content
    - compression: enabled_for_all_responses
    - regional_data_processing: keep_data_close_to_users

reserved_capacity:
  rds_reserved_instances: 1_year_term
  elasticache_reserved_nodes: 1_year_term
  savings_plans: compute_savings_plan_1_year

tagging_strategy:
  mandatory_tags:
    - Environment: [production, staging, development]
    - Service: [web, worker, database, cache]
    - Owner: [team-name]
    - Project: jobhive
    - CostCenter: engineering
    - Backup: [required, not-required]
    -
  optional_tags:
    - Version: application_version
    - Component: [api, frontend, ml-engine]
    - Schedule: [24x7, business-hours, on-demand]

performance_tuning:
  database:
    - connection_pooling: pgbouncer_sidecar
    - query_optimization: automated_explain_plans
    - index_optimization: quarterly_review
    - read_replicas: 2_replicas_for_analytics
    
  caching:
    - redis_cluster: 3_nodes_with_replication
    - application_cache: django_cache_framework
    - cdn_cache: cloudfront_edge_caching
    - browser_cache: appropriate_cache_headers
    
  application:
    - async_processing: celery_task_queue
    - static_compression: gzip_and_brotli
    - image_optimization: automated_resizing
    - api_response_compression: enabled

performance_metrics:
  application_metrics:
    - response_time_p95: target_200ms
    - throughput: requests_per_second
    - error_rate: target_below_0.1%
    - availability: target_99.9%
    
  infrastructure_metrics:
    - cpu_utilization: target_below_70%
    - memory_utilization: target_below_80%
    - disk_io: iops_and_throughput
    - network_latency: cross_az_communication
    
  business_metrics:
    - interview_completion_rate: target_above_95%
    - user_satisfaction: response_time_perception
    - conversion_funnel: signup_to_active_user
    - revenue_impact: performance_correlation