# Product Catalog Architecture (50M Scale) ## Data Model & Hierarchy The catalog supports up to **50,000,000 SKUs** utilizing 26-character ULID primary keys. ```mermaid erDiagram Category ||--o{ Product : contains Brand ||--o{ Product : manufactures DeviceSeries ||--o{ Product : applies_to DeviceModel ||--o{ Product : applies_to Product ||--|{ ProductVariant : has ProductVariant ||--|{ VariantAttribute : defines AttributeType ||--|{ VariantAttribute : types ``` --- ## High-Scale Pagination Strategy For scaling beyond 100,000 items, standard SQL `OFFSET` pagination (e.g. `LIMIT 50 OFFSET 100000`) becomes prohibitively slow ($O(N)$ scan). ### Keyset / Cursor-Based Pagination All catalog API endpoints enforce Keyset Cursor Pagination using `created_at` and `product_id`: ```sql SELECT product_id, name, price, created_at FROM products WHERE (created_at, product_id) < (:last_created_at, :last_product_id) ORDER BY created_at DESC, product_id DESC LIMIT 50; ``` --- ## Outbox Catalog Indexing Pipeline When products or variants are updated in MySQL: 1. An event is written to `outbox_events` (`aggregate_type='PRODUCT'`, `aggregate_id=product_id`, `aggregate_version=v42`). 2. Celery Worker picks up the event, reads `SearchService`, and updates OpenSearch asynchronously.