Milvus Terminology

  • Collection: A collection that consists of a set of entities and are equivalent to a table in an RDBMS.

  • Segment: A data file that Milvus automatically creates by merging inserted data. A collection can contain multiple segments. One segment can contain multiple entities. During search, Milvus searches each segment, filters deleted data, and returns the combined result.

  • Entity: A group of fields that correspond to real world objects. These fields can be structured data representing object properties or vectors representing object features.

  • Entity ID: A guaranteed unique value that can be used to always reference an entity.

    Currently, Milvus does not support entity ID de-duplication and it is possible to have duplicate IDs in a segment.
  • Field: A field within an entity. A field can either be structured data, such as numbers, strings, or unstructured data, such as vectors.

  • Vector: A type of field representing the feature of an object.

    Currently, an entity can only contain up to one vector.
  • Index: An index built based on raw data and improves the speed of data retrieval operations on a collection.

  • Mapping: A set of rules that define how data is organized in a collection.

On this page