Aurora

(higher-scale relational + serverless options)

Mental model (when it wins)

• “RDS, but engineered for higher throughput + read scaling + storage decoupling.”
• Great for: high read fanout, multi-AZ fast recovery patterns, Global Database, and sometimes bursty relational workloads.

The knobs that matter

• Aurora Standard vs Aurora I/O-Optimized

  • Standard: pay storage + I/O requests.
  • I/O-Optimized: higher instance/storage rates, $0 I/O (wins when I/O heavy). ([Amazon Web Services, Inc.][4])

• Serverless v2

  • Capacity in ACUs; 1 ACU ≈ ~2 GiB RAM (rule of thumb). ([AWS Documentation][5])
  • Set min/max ACU to control spend and spikes.

• Readers: add reader instances to scale reads.

• Data API: convenient, but is its own priced surface; use intentionally. ([Amazon Web Services, Inc.][4])

Pricing mental model (back-of-envelope)

Aurora cost = instance-hours + storage GB-month + (I/O requests if Standard) + replication/data transfer for global patterns. ([Amazon Web Services, Inc.][4]) From AWS examples:

• Storage often modeled around $0.10/GB-month (Standard example), and I/O around $0.20 per 1M I/Os for Standard. ([Amazon Web Services, Inc.][4]) Serverless v2 uses ACU-hours (example shows $/ACU-hour). ([Amazon Web Services, Inc.][6])

Senior heuristic: If your workload is I/O heavy (lots of small reads/writes), model I/O-Optimized—it can be cheaper overall. ([Amazon Web Services, Inc.][4])

Terraform template (Aurora PostgreSQL Serverless v2)

resource "aws_rds_cluster" "aurora" {
  cluster_identifier = "${var.name}-aurora"
  engine             = "aurora-postgresql"
  engine_version     = var.engine_version

  database_name   = var.db_name
  master_username = var.username
  master_password = var.password

  db_subnet_group_name   = var.db_subnet_group_name
  vpc_security_group_ids = [var.db_sg_id]

  backup_retention_period = 7
  deletion_protection     = true

  serverlessv2_scaling_configuration {
    min_capacity = 0.5
    max_capacity = 8
  }

  tags = var.tags
}

resource "aws_rds_cluster_instance" "writer" {
  identifier         = "${var.name}-aurora-writer"
  cluster_identifier = aws_rds_cluster.aurora.id
  instance_class     = "db.serverless"
  engine             = aws_rds_cluster.aurora.engine
  engine_version     = aws_rds_cluster.aurora.engine_version
}

variable "name"                { type = string }
variable "engine_version"      { type = string default = "16.1" }
variable "db_name"             { type = string default = "app" }
variable "username"            { type = string }
variable "password"            { type = string sensitive = true }
variable "db_subnet_group_name"{ type = string }
variable "db_sg_id"            { type = string }
variable "tags"                { type = map(string) default = {} }

---