Explaining AWS DynamoDB
2 min readAug 26, 2024
- Let’s start personally. Werner Voegles is a badass, passionate technologist, loving EDM, CTO of Amazon, blogger at https://www.allthingsdistributed.com/, and video-podcaster of “Now Go Build!” docuseries on solving problems with engineering.
- Vogels described the deep technical nature of Amazon’s infrastructure work in a paper about Amazon’s Dynamo, the storage engine for Amazon’s shopping cart.
- In 2007, he wrote an influencing paper describing Dynamo — Amazon’s Highly Available Key-Value Store (https://www.allthingsdistributed.com/files/amazon-dynamo-sosp2007.pdf)
- This is a blueprint of what we now call DynamoDB (DDB), and an important milestone in the early history of the cloud
- DDB is a managed multi-AZ NoSQL datastore with cross-region replication to keep data safe
- I.e. it is High Availability (HA)
- It takes a bit to distribute all rights across all AZ → term “eventual consistency”; usually less than a second
- This means that is you make a write and then immediately after a read — you reflect the old data;
- If you really need reliably latest data, you can request a strongly consistent read; the downside is that the read may fail in some conditions
- DDB is usually BASE-compliant as opposed to ACID.
- There is possibility of ACID compliancy (DDB transactions opening it up to RDBMS capabilities)
- DDB is priced on throughput, usually
- You can use autoscalijng per configured min/max capacities.
- You can use on-demand capacity, there is a small premium price for that — if you don’t know what read/capacity should be
More generally: Relational VS NoSQL DBs
- Tables and fields support relationships between them
- NoSQL record is usually self-contained and does not have to relate to any other tables to make sense; the hierarchy is there
- Together they are called attributes, and many of them form a table
- Every record must have a unique identifier, in DDB this is called a primary key
- DDB creates an internal hash, to decide to store the value, it’s also called hash attribute
- You can key with two attributes, with partition key VS sort key
