Database
Amazon DynamoDB
Serverless, fully managed NoSQL database with key-value and document data models, delivering single-digit millisecond performance at any scale. It handles operational workloads that need consistent performance without managing servers or the complexity of scaling a relational database.
Exam phrases that point here
«single-digit millisecond latency at any scale»
DynamoDB is a serverless NoSQL key-value and document database, designed to respond in single-digit milliseconds even as traffic and data grow.
When not: If the scenario requires joins between tables, complex SQL queries, or referential integrity, it points to a relational database such as Amazon RDS or Amazon Aurora. And if it asks for microseconds, the clue is DAX in front of DynamoDB.
«unpredictable traffic and you do not want to plan capacity»
On-demand capacity mode charges per request and adjusts to traffic on its own, without having to estimate reads and writes per second.
When not: If traffic is steady and can be forecast, provisioned mode fits (normally with auto scaling), which gives more predictable costs. And if the application needs SQL, a serverless relational option such as Amazon Aurora Serverless is a better fit.
«microsecond response times for reads on a DynamoDB table»
DynamoDB Accelerator (DAX) is an in-memory cache compatible with the DynamoDB API: it brings reads down from milliseconds to microseconds with minimal changes to the application.
When not: DAX is not suitable if the application needs strongly consistent reads, nor is it a good fit for write-heavy workloads. If the source database is relational, the cache to think of is Amazon ElastiCache.
«run code automatically every time an item in the table changes»
DynamoDB Streams records every insert, update, or delete in order, and an AWS Lambda function can be triggered with those records.
When not: The stream keeps changes for only 24 hours. If you need longer retention, more consumers, or integration with other streaming tools, Kinesis Data Streams for DynamoDB fits. If events come from many different services, think of Amazon EventBridge.
«local reads and writes in multiple Regions, active-active»
Global tables replicate the table in the Regions you choose and all of them accept reads and writes; there is no primary Region.
When not: If the database is relational, the answer is usually Amazon Aurora Global Database, which writes in a primary Region. If the application uses Cassandra, Amazon Keyspaces offers multi-Region replication.
«automatically delete expired items without consuming write capacity»
Time to Live (TTL) deletes each item when the timestamp it stores expires, without using write capacity. It is useful for sessions, temporary data, or records that lose value.
When not: If the goal is to keep sessions in memory with sub-millisecond latency, Amazon ElastiCache also fits. If you are dealing with objects in S3, age-based deletion is done with S3 lifecycle rules, not TTL.
Often paired with
- AWS Lambda