Elasticsearch产品目录存储：单文档与多索引方案选型咨询

Based on the product catalog needs and search requirements you've laid out, storing each product as a single document in a single Elasticsearch index is absolutely your optimal choice—no need to split into multiple associated indexes. Here's why:

Why single-document storage works perfectly for your needs #

• Seamless support for core search features: Your requirements (paginated product lists, sorting by brand/title/price/discount, filtered by pre-configured parameters like root category) are exactly what Elasticsearch excels at with single-document models. You can map fields like brand, root_category as keyword types for fast filtering, price/discount as numeric types for efficient sorting, and title as a text type for full-text search. Pagination will be straightforward and performant without any cross-index overhead.
• Avoids the performance pitfalls of cross-index joins: Elasticsearch's nested or parent-child relationships (used for cross-index associations) introduce significant performance costs, especially for pagination and sorting at scale. Joining data across indexes complicates queries, slows down response times, and can even lead to inconsistent pagination results—none of which you need here since all your filters and sort criteria are tied directly to the product itself.
• Simplifies data maintenance: With a single document per product, all attributes live in one place. Updating a product's price, discount, or category only requires modifying one document, eliminating the risk of data sync issues between multiple indexes.

When would splitting into multiple indexes make sense? #

Just to give you context, you'd only need to consider multi-index setups if:

• You have highly dynamic, frequently updated associated data (like real-time inventory levels from a separate system, or user-generated reviews that change constantly) that you don't want to reindex the entire product document for every update.
• You need complex cross-entity aggregations or queries that involve joining products with entirely separate entities (like suppliers, warehouses, or multi-level category hierarchies with their own independent metadata).

But neither of these scenarios apply to your stated requirements, so sticking with single-document storage is the way to go.

内容的提问来源于stack exchange，提问作者marek