Compare databases on AWS

December 1, 2022

This post is my note taken from the Udemy course Ultimate AWS Certified Solutions Architect Associate SAA-C03 instructed by Stephane Maarek.

There are a number of database services out there on AWS. Choosing them correctly is related to a number of dimensions:

The characteristic of your data throughput (read-heavy or write-heavy, does the traffic load fluctuate etc.)
The size of your data and desired retention period
What the source of truth for the data is
Latency and concurrency requirement
Data schema requirement
License cost

Depending on those different factors, AWS generally provides these options:

RDBMS (Relational Database Management System) = SQL / OLTP, such as RDS or Aurora
NoSQL databases, such as DynamoDB, ElastiCache
Object Store, such as S3
Data Warehouse, such as Redshift (or OLAP), Athena, EMR
Search type, such as OpenSearch
Graph type: such as Amazon Neptune
Ledger type: such as Amazon Quantum Ledger Database
Time series type: such as Amazon Timestream

General comparison

Database	Type	Use Case
RDS	Managed SQL DB (PostgresSQL, MySQL ec.)	Relational DB
Aurora	PostgresSQL / MySQL compatible Amazon SQL DB	Same as RDS but less maintenance
ElastiCache	Managed Redis or Memcached	In-memory data store with extremely low latency
DynamoDB	Amazon Proprietary NoSQL DB	Great to rapidly evolve schemas
DocumentDB	Amazon version of MongoDB	Similar with DynamoDB, but in MongoDB norm
S3	Key value object storage	Static files, big files, website hosting
Neptune	Amazon graph database	Fraud Detection, recommendation, social network…
QLDB	Managed serverless immutable DB	review change history, good for financial regulation
Keyspaces	Managed serverless version of Apache Cassandra (NoSQL)	IoT devices info, time-series data
Timestream	Managed serverless time series database	Trillions events per day, IoT, real-time analytics…

All databases above support auto-scale, though their auto-scale capability varies. Some scale with limitation, while some others scale infinitely
All databases above support Multi-AZ, but S3 provides a One Zone-IA option which only stores data in one availability zone

Requires provisioning beforehand
Automated backup with Point in Time restore up to 35 days
Manual DB snapshot for longer-term recovery
IAM authentication supported
Managed maintenance (to patch the underlying instances) with downtime, can be scheduled
The RDS Custom option is available for Oracle and SQL Server, which allows user accessing the underlying instances (not allowed for other RDS types)
Performance: RDS Proxy to reuse DB connection and decrease the fail-over time

Mostly same with RDS, but only compatible with PostgresSQL and MySQL
Storage and compute are separated
Same security, monitoring and maintenance feature as RDS
For unpredictable workloads, there is an option Aurora Serverless
For continuous writes fail-over, Aurora Multi-Master
For global availability, Aurora Global (up to 16 read instances in each region and < 1s replication)
To perform ML using SageMaker and Comprehended on Aurora, Aurora Machine Learning
To replicate data, Aurora Database Cloning

With TTL feature, DynamoDB can replace ElastiCache as a key/value store
Millisecond latency, DAX cluster (read cache) further reduces the read latency to microsecond level
Event Process streaming: DynamoDB Streams or Kinesis Data Streams
Global Table feature: active-active
Automated backup up to 35 days with PITR (restore to new table)
Manual backup for longer-term recovery
Exporting and importing to and from S3 is supported within the PITR window. This won’t cost the RCU or WCU
No SQL query language supported

Good for bigger objects, not so good for many small objects
Special features: versioning, encryption, replication, MFA delete, access logs
Security control: IAM, bucket policy, ACL, access points, object lambda, CORS, obejct/value lock
Encryption: SSE-S3, SSE-KMS, SSE-C, client-side, TLS in transit and optional default encryption
Batch operations supported
Performance: multi-part upload, S3 Transfer Acceleration, S3 Select
Automation: S3 Event Notification (SNS, SQS, Lambda, EventBridge)