# Hot partition — back-of-envelope (DynamoDB / Kinesis mental model)

**Not a substitute for AWS calculator or Support.** Use to sanity-check whether a single partition key can carry your peak write RPS.

Given:

- `W` = sustained writes per second you need on one logical key (e.g. tenant id `T1`)
- DynamoDB softly budgets ~1000 RCU/WCU per partition for planning discussions (actual throttling depends on burst, adaptive capacity, on-demand mechanics — verify in docs for your region and table class)

Ask: if every write hits the **same** sort key prefix or the **same** Kinesis partition key, what breaks first?

Document assumptions and re-run in **Provisioned** vs **On-demand** modes separately. Link capacity findings to [single-table design posts](https://www.factualminds.com/blog/dynamodb-single-table-design-patterns-for-saas/) when you outgrow naive keying.
