Queries Before Tables

A team ports a product catalog to DynamoDB by translating each relational table into its own DynamoDB table, one per entity. Rendering a single product page now costs seven separate requests. What did the design skip?

Adding a secondary index to every table, so each entity can also be fetched by its name

Listing the read patterns first, then shaping keys so one query returns everything a page needs

Normalizing further, so each entity table stays small enough to scan cheaply per request

Provisioning higher throughput, so the seven requests finish inside the page latency budget