DMS

🧠 What is AWS DMS?

AWS Database Migration Service (DMS) helps you migrate databases from on-premises, cloud, or hybrid environments to AWS, between AWS services, or even from AWS to other clouds.

✅ It supports both homogeneous (e.g., MySQL to MySQL) and heterogeneous (e.g., Oracle to Aurora) migrations with near-zero downtime.

---

🚀 Common Use Cases

Use Case	Example
Cloud adoption	Migrate Oracle from on-prem to Amazon RDS
Cross-engine migration	Migrate SQL Server to Aurora PostgreSQL
Continuous replication	Sync data from on-prem to AWS in real-time
Analytics	Migrate transactional DB to Amazon Redshift
Disaster recovery	Maintain DR copy in another region

---

🔁 DMS Supports

Source DBs	Target DBs
MySQL, PostgreSQL	RDS (Aurora, MySQL, PostgreSQL, etc.)
Oracle, SQL Server	Amazon S3, Redshift, DynamoDB
MariaDB, MongoDB	Kinesis, OpenSearch, DocumentDB
SAP ASE, IBM DB2, etc.	Snowflake (via Kinesis or S3)

---

⚙️ How DMS Works

Basic Components:

1. Replication Instance
   → EC2-based instance that runs the migration

2. Source Endpoint
   → Connection configuration to your source database

3. Target Endpoint
   → Connection configuration to your target database

4. Migration Task
   → Defines what to move, how, and with what rules

---

🔄 Migration Modes

Mode	Description
Full Load	One-time copy of existing data
CDC (Change Data Capture)	Replicate ongoing changes
Full Load + CDC	Initial data load + continuous sync

🔄 CDC uses the source DB's transaction logs to replicate changes.

---

📋 DMS Migration Flow

Source DB → DMS Replication Instance → Target DB
                  |
       (via VPC, IAM, Endpoints, KMS)

---

🧪 Example Terraform Snippet for DMS

resource "aws_dms_replication_instance" "example" {
  replication_instance_id = "example-dms-instance"
  replication_instance_class = "dms.t3.micro"
  allocated_storage = 50
}

resource "aws_dms_endpoint" "source" {
  endpoint_id = "source-endpoint"
  endpoint_type = "source"
  engine_name   = "mysql"
  username      = "admin"
  password      = "password"
  server_name   = "source-db.mycompany.com"
  port          = 3306
}

resource "aws_dms_endpoint" "target" {
  endpoint_id = "target-endpoint"
  endpoint_type = "target"
  engine_name   = "aurora"
  username      = "admin"
  password      = "password"
  server_name   = "aurora-cluster.endpoint.rds.amazonaws.com"
  port          = 3306
}

resource "aws_dms_replication_task" "task" {
  replication_task_id = "my-task"
  replication_instance_arn = aws_dms_replication_instance.example.replication_instance_arn
  source_endpoint_arn = aws_dms_endpoint.source.endpoint_arn
  target_endpoint_arn = aws_dms_endpoint.target.endpoint_arn
  migration_type = "full-load-and-cdc"
  table_mappings = <<EOF
{
  "rules": [
    {
      "rule-type": "selection",
      "rule-id": "1",
      "rule-name": "1",
      "object-locator": {
        "schema-name": "%",
        "table-name": "%"
      },
      "rule-action": "include"
    }
  ]
}
EOF
}

---

🧾 Pricing (2024)

Resource	Cost
Replication instance	Starts at ~$0.018/hr (dms.t3.micro)
Data transfer	Free between AWS services in same region
Storage (logs/checkpoints)	Based on allocated volume
Free tier	750 hours/month for 2 months (for t3.micro)

---

🔐 Security

Feature	Supported
IAM Role-based access	✅ Yes
KMS Encryption	✅ Yes (at rest)
SSL for DB Connections	✅ Yes
VPC/Private Subnet	✅ Yes (best practice)
CloudWatch & SNS	✅ Logs + alerts

---

🛠️ Best Practices

• Place replication instance in same region and subnet as target

• Enable multi-AZ for replication if needed

• Use parameter groups to adjust retry/timeouts

• Pre-create target schema (DMS won’t create them in all engines)

• Monitor task health using CloudWatch metrics and logs

---

🚨 Limitations

Area	Details
Schema conversion	Not automatic → Use AWS SCT (Schema Conversion Tool)
Triggers, SPs	Not migrated; only data + indexes
Data types	Some conversions require mapping (e.g., BLOB to JSON)
CDC dependencies	CDC needs binary logs (e.g., binlog for MySQL)

---

✅ TL;DR Summary

Feature	AWS DMS
Migration types	Full Load, CDC, or both
Source/Target support	SQL, NoSQL, S3, Redshift, etc.
Real-time replication	✅ Yes (CDC)
Fully managed	✅ Yes
Terraform support	✅ Yes (with full component coverage)
Free tier	✅ 2 months (750 hours on t3.micro)
Schema migration	❌ Use AWS SCT separately

---