> For the complete documentation index, see [llms.txt](https://mamawhocode.gitbook.io/aws/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://mamawhocode.gitbook.io/aws/services/database/rds.md).

# RDS

Relational Database Service

[FAQs](https://aws.amazon.com/rds/faqs/?saa=sec\&sec=prep) |

## Overview

![Amazon RDS](https://aws.dendron.so/assets/images/Amazon-RDS_en.jpg)

* a `managed` DBaaS service (auto OS patching, DB update, backups...) from AWS.

***

## Features

* Supports 6 RDS database engines (MariaDB, MySQl, PostgresSQL, SQL Server, Oracle, Aurora (PostgreSQl, MySQL))
* 3 types of instance classes
  * Standard (m1\~m6)
  * Memory optimized (r5, x1, z1d, r6g...)
  * Burstable performance (t2, t3)
* 3 types of storage
  * General purpose (SSD) storage. Eg: gp2
  * Provisioned IOPS (SSD) storage. For I/O intensive workload
  * Magnetic storage. For backward compatibility.
* optimized for memory, performance or I/O.
* High availability: [Multi-AZ](#multi-az-vs-read-replica) and Fail-over.
* `Point-in-time` restore
  * Automatically back up
  * Daily snpashots and transaction logs.

### Enhanced Monitoring

* A feature *<mark style="color:red;">available specifically for RDS</mark>*.
* Provides detailed metrics about the health of your RDS instances, giving more insight into database load and performance.&#x20;

-> Particularly useful for detailed analysis, like how different processes use the CPU.

#### vs. Cloud Watch metrics

* CloudWatch collects CPU utilization metrics from the hypervisor.
* ***Enhanced Monitoring*** collects metrics directly *<mark style="color:red;">from an agent</mark>* on the DB instance.

### Custom endpoint

For different works you need different endpoint. Some works are important, some are not (data analysis, report).\
-> Make sure important tasks are done quickly, without being slowed down by the less important ones.

### Scalability

auto scaling when

* running of storage
  * free storage less than 10%
  * low-storage lass at least 5 minutes
  * 6 hours have passed since last modification
* you have to set `Maximum storage threshold` (maximum limit for the storage).

***

## Pricing

Depend on many factors

* Instance type
  * On-demand
  * Reserved instance (1 or 3 years term)
* Stop/Start instance (the underlying is EC2 instance)
* Storage capacity
* Data transfer
* Additional features

***

## Security

<figure><img src="https://2259236002-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fuh9xZDZ53qGqmMCM44PU%2Fuploads%2Fu4ypXUjyUWwkaNCh3eUN%2Fimage.png?alt=media&amp;token=a4f372f8-112f-45be-b457-2937d61d999a" alt=""><figcaption><p>3 layers of network security</p></figcaption></figure>

* Using 3 layer of security
  * Isolated VPC
  * Network ACLs (subnet level)
  * Security group (instance level)
* Resource-level security: using IAM
* Encryption at rest and in transit
  * Encryption at rest (256-bit AES algorithms) using AWS KMS.
  * Encrypt data before store on disk using Transparent data encryption (TDE).
  * Encryption in transit. Eg: connection to DB are secured using HTTPS, using IPsec VPN.

***

## RDS Features

### RDS Custom

* Managed Oracle and SQL Server Database with OS and database customization.
* De-activate Automation Mode.
* Benefits
  * all the benefits of a managed service
  * access to te underlying database and OS as you can.
    * config setting
    * install patches
    * enable native features
    * access underlying EC2 instances using `SSH` or SSM Session Manager

### RDS Backups

* Automated backups (Automated backup turn ON by default)
  * Daily full backup.&#x20;
  * Transaction log backed-up every 5 minutes (to S3)
* Manual DB snapshot
  * Manually triggered by user
  * Retention of backup for as long as you want

### RDS & Aurora Security

* At-rest encryption
  * using AWS KMS - must be defined as lauch time
* If the master is not encrypted, the read replicas can not be encrypted either.
* To encrypt an un-encrypted database
  * 1st, create a snapshot
  * 2nd, restore that DB from a snapshot as encrypted.

### Aurora multi-master cluster

* Support both read/write heavy workload. (Not like Read-replica, which only support read-heavy workload)
* Unlike Read-Replica, which has **Master** and **Secondary** instance, multi-master cluster can take any instance to read/write as a Master.

<figure><img src="https://2259236002-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fuh9xZDZ53qGqmMCM44PU%2Fuploads%2FIBeX1MAolktMgMIIZjHv%2Fimage.png?alt=media&amp;token=d228c9b4-205b-4e20-ba59-cb19efc9ceef" alt=""><figcaption><p>Aurora multi-master cluster</p></figcaption></figure>

### RDS Proxy

* Fully managed database proxy for RDS
* It helps *<mark style="color:red;">lower the connection number</mark>* go through the DB instance.
* Never publicly accessiable (must be accessed from VPC)
* Benefit
  * improving DB efficiency
  * serverless, autoscaling, HA (multi-AZ)
  * supports all RDS & Aurora
  * reduced RDS & Aurora failover time by up 66%

## Multi-AZ vs Read Replica

* Multi-AZ standby instance

  * For **disaster recovery / failover** (auto fail over to Standy instance). Increase `availability`.
  * <mark style="color:red;">SYNC</mark> replication

  <figure><img src="https://2259236002-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fuh9xZDZ53qGqmMCM44PU%2Fuploads%2FYW28brJ7ckv96GLc2nCc%2Fimage.png?alt=media&amp;token=03bc8e23-cbb7-4cbf-8d5a-32c20e630c4a" alt=""><figcaption><p>turn on Multi-AZ in RDS</p></figcaption></figure>
* Read- Replica (for Read-heavy workload)
  * For **READ-scalability**
  * <mark style="color:red;">ASYNC</mark> replication
  * The Read replica DB can be aligned in the same AZ, cross AZ or cross Region
  * Free within the same region

<figure><img src="https://2259236002-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fuh9xZDZ53qGqmMCM44PU%2Fuploads%2FImsGatGKPgt55Hc0Kfsz%2Fimage.png?alt=media&amp;token=e4f736d6-013e-4e8e-bc94-7aaa3019c8a2" alt=""><figcaption><p>Read-replica</p></figcaption></figure>

***

## Trivia

* Maximum of CPU is 32 vCPUs, and maximum of memory is 244 GiB RAM.
* Storage scaling depend on type of engine
  * Aurora (auto scaling): max is 64TB, or your defined number.
  * SQL Server: 16TB
  * Other (Oracle, and 3 open-sources): 64TB
* Encrypted at rest include: Automated backups, Read replicas, and Snapshots.
* When you turn ON Multi-AZ, AWS *automatically* creates a standby instance in a different AZ.
* By stopping the DB although you are not paying for DB hours you are still paying for `Provisioned IOPs`
* Do not confuse [Enhanced Monitoring](#enhanced-monitoring) feature with [Detailed Monitoring](/aws/services/compute/ec2.md#features) of EC2.