How to model
mapping design is very important and need to be considered from two dimensions:
- Features: search, sort, aggregate
- Performance: The performance of the storage unlock, memory overhead, search
mapping Note:
- It is easy to add a new field (required update_by_query if necessary)
- Update Delete field does not allow (reconstruction needs reindex data)
Best Practices
1, does not allow automatic new field, the dynamic set to strict. The default is true;
2, no word field, the type provided keyword. Use the default multi-field characteristics, text, keyword these two types have;
3, no need to check the field, the index set to false. The default is true;
4, and does not need to sort the polymerization field, doc_values arranged false. The default is true;
5, no checking, sorting, polymeric field, arranged to enable false, do only memory;
6, type = text field can not sort by default, such as the need to sort the fielddata set to true, the default is false;
7, to avoid excessive single index field, the default maximum value of 1000;
8, to avoid the problem of inaccurate polymerization caused by nulls;
9. Avoid using regular query;
10, try not to associate index designed to be a little more redundant fields, space for time, truthfully can not be avoided, according to the following ways: