This bookmark provides specialized views for analyzing the direct memory consumption of the various stores of the application, with a focus on the fields the chunks can be attributed to.
This view lists the most memory-consuming fields across all stores.
This page only aims to be an entry-point for further analysis, after drawing attention to abnormally high memory consumptions.
In the case of a name conflict between fields of different stores, the memory consumption is displayed per owner under each field.
Chunks may be shared across multiple fields (in the case of indexes over multiple fields for example). As such, a very large index will impact the memory consumption of all the fields it is attributed to.
This view displays the per-component memory consumption of each field per store.
The various possible components that can be attributed to a field are:
-
INDEX
: an index indexes the corresponding field. It is usually the highest direct-memory-consuming component for store fields.All indices indexing this field are taken into account (e.g. an index on fields
A, B
and another index on fieldB
both contribute to field B's memory consumption)An index attributes its chunks to all fields it indexes.
-
DICTIONARY
-
RECORDS
: the chunks the stores use to store their fields' values fall into this category. -
VECTOR_BLOCK
: specific to vector fields, the memory consumption of the corresponding vector blocks (includes swapped memory)
It is important to note that not all chunks of a store appear in this page.
Some store chunks are not be attributed to any field in particular:
- the chunks of the versioning column of a store (
RECORDS
component) - partition mapping chunks used by references between stores (
REFERENCE
component)