Databases

DynamoDB

NoSQL database that is fully managed with massive scale, with no disk space to provision and a max object size of 400KB. Allowing for backups and point in time recovery, and standard and infrequent access types.

DynamoDB Limitations

must have a primary key for each item, or a composite key index; cannot query by anything other than primary key + sort key (composite key)

DynamoDB Streams

react to changes in DynamoDB tables in real time, can be Lambda, EC2, etc. with 24 hours of data retention

DynamoDB Global Tables

cross-region replication, that must enable DynamoDB streams to be able to use, good for DR and low latency, an active/active type of replication

DynamoDB Accelerator (DAX)

seamless cache for DynamoDB, requiring no rewrite of application with microsecond latency for cached operations, solving the hot-key problem with a default 5 minute TTL. Cannot be used for computations, only queries.

OpenSearch

provides searing and indexing capabilities for any database (e.g. DynamoDB) that can be managed or serverless clusters

Logstash

log ingestion mechanism in OpenSearch similar to CW Logs

Relational Database Service (RDS)

a managed database service (e.g. PostgreSQL, MySQL, Oracle, IBM) that automates provisioning, backups, patching and monitoring, launched within a VPC controlling access using security groups. Storage is backed by EBS and can increase with autoscaling. Backups are automated but do have a expiration date.

RDS Events

get notified of all events in SNS (e.g. operations, outages, backups)

RDS Multi-AZ Strategy

standby instance in case of failover that is synced from one the active instance (active/passive).

RDS Read Replicas

cross-origin snapshots that are backed up into read-replicas of the main database, cannot write in the replicas.

RDS Distributed reads

weighted RDS record set that will distribute requests to each replica, with health checks to only include healthy RDS databases

RDS Security

KMS encryption of EBS snapshots, with Transparent Data Encryption (TDE) for Oracle and SQL Server, SSL for all databases, and IAM for PostgreSQL, MySQL, and MariaDB. CloudTrail cannot be used to track queries made in RDS

RDS Oracle Backups

RMAN (recovery manager) to restore non-RDS native databases (upload to an S3 bucket -> DB), Read Application Clusters are not supported (only on EC2)

RDS MySQL Backups

use the SQL Dump tool to migrate a native DB to a non-native database, can be an on-prem or an EC2 instance.

RDS Proxy for Lambda

opens and maintains a database connection in a Lambda function, without needing to handle connections/pools, with support for IAM roles, autoscaling and native database authentication. Lambda must have connectivity to be able to use.

Aurora

PostgreSQL and MySQL managed database that automatically grows up to 128TB, with 6 copies of data in multiple availability zones. Self-healing with peer-to-peer replication, spread across hundreds of volumes, automatically scaling at 10GB at a time.

Aurora Backups

backups and snapshot and restore, the same as RDS. Can offload data to/from S3 for efficient use of resources .

Aurora Cross Region Replica

copy of the entire database in another region, cannot select target tables to replica.

Aurora DB Cluster

shared, auto-expending storage volume with a writer endpoint and read replicas from the shared storage volume.

Aurora Cluster Endpoint

writer endpoint, connecting to primary DB instance and used for write operations

Aurora Reader Endpoint

load balancing for read-only connections to replicas

Aurora Custom Endpoint

set of select DB instances in the cluster that will be connected to endpoint, with different capacities and configurations.

Aurora Instance Endpoint

connects to a specific DB instance in the cluster, used for diagnostics and fine-tuning for a specific instance.

Aurora Logs

showing errors, slow queries, general and audit information, can be downloaded or published in CloudWatch

Performance Insights Tool

find issues by waits, SQL statements, hosts, and users for RDS and Aurora

Aurora Data API

secure HTTPs connection to Aurora database that allows SQL statements to be run without a jdbc connection, that is non-persistent and must have permissions to see credentials in Secrets Manager

RDS Proxy for Aurora

create a read-only endpoint that connects to Aurora read-replicas only (from a Lambda)

Global Aurora

cross region replication, that has a primary region, and multiple secondary regions with a < 1s replication lag, with up to 16 replicas per region for better RTO and RPO

Aurora Write Forwarding

enables secondary DB clusters in Global Aurora architecture to forward SQL statements to the main database cluster, then is replicated to the cluster that the SQL originated from.

RDS to Aurora Conversion

create an RDS snapshot and restore to Aurora, not great for latest data, or create an Aurora Read Replica to the RDS RB.