@@ -40,94 +40,3 @@ better. For instance if a user searches for two words `foo` and `bar`, a match
4040across different chapters is probably very poor, while a match within the same
4141paragraph is likely good.
4242
43- [float]
44- [[sparsity]]
45- === Avoid sparsity
46-
47- The data-structures behind Lucene, which Elasticsearch relies on in order to
48- index and store data, work best with dense data, ie. when all documents have the
49- same fields. This is especially true for fields that have norms enabled (which
50- is the case for `text` fields by default) or doc values enabled (which is the
51- case for numerics, `date`, `ip` and `keyword` by default).
52-
53- The reason is that Lucene internally identifies documents with so-called doc
54- ids, which are integers between 0 and the total number of documents in the
55- index. These doc ids are used for communication between the internal APIs of
56- Lucene: for instance searching on a term with a `match` query produces an
57- iterator of doc ids, and these doc ids are then used to retrieve the value of
58- the `norm` in order to compute a score for these documents. The way this `norm`
59- lookup is implemented currently is by reserving one byte for each document.
60- The `norm` value for a given doc id can then be retrieved by reading the
61- byte at index `doc_id`. While this is very efficient and helps Lucene quickly
62- have access to the `norm` values of every document, this has the drawback that
63- documents that do not have a value will also require one byte of storage.
64-
65- In practice, this means that if an index has `M` documents, norms will require
66- `M` bytes of storage *per field*, even for fields that only appear in a small
67- fraction of the documents of the index. Although slightly more complex with doc
68- values due to the fact that doc values have multiple ways that they can be
69- encoded depending on the type of field and on the actual data that the field
70- stores, the problem is very similar. In case you wonder: `fielddata`, which was
71- used in Elasticsearch pre-2.0 before being replaced with doc values, also
72- suffered from this issue, except that the impact was only on the memory
73- footprint since `fielddata` was not explicitly materialized on disk.
74-
75- Note that even though the most notable impact of sparsity is on storage
76- requirements, it also has an impact on indexing speed and search speed since
77- these bytes for documents that do not have a field still need to be written
78- at index time and skipped over at search time.
79-
80- It is totally fine to have a minority of sparse fields in an index. But beware
81- that if sparsity becomes the rule rather than the exception, then the index
82- will not be as efficient as it could be.
83-
84- This section mostly focused on `norms` and `doc values` because those are the
85- two features that are most affected by sparsity. Sparsity also affect the
86- efficiency of the inverted index (used to index `text`/`keyword` fields) and
87- dimensional points (used to index `geo_point` and numerics) but to a lesser
88- extent.
89-
90- Here are some recommendations that can help avoid sparsity:
91-
92- [float]
93- ==== Avoid putting unrelated data in the same index
94-
95- You should avoid putting documents that have totally different structures into
96- the same index in order to avoid sparsity. It is often better to put these
97- documents into different indices, you could also consider giving fewer shards
98- to these smaller indices since they will contain fewer documents overall.
99-
100- Note that this advice does not apply in the case that you need to use
101- parent/child relations between your documents since this feature is only
102- supported on documents that live in the same index.
103-
104- [float]
105- ==== Normalize document structures
106-
107- Even if you really need to put different kinds of documents in the same index,
108- maybe there are opportunities to reduce sparsity. For instance if all documents
109- in the index have a timestamp field but some call it `timestamp` and others
110- call it `creation_date`, it would help to rename it so that all documents have
111- the same field name for the same data.
112-
113- [float]
114- ==== Avoid types
115-
116- Types might sound like a good way to store multiple tenants in a single index.
117- They are not: given that types store everything in a single index, having
118- multiple types that have different fields in a single index will also cause
119- problems due to sparsity as described above. If your types do not have very
120- similar mappings, you might want to consider moving them to a dedicated index.
121-
122- [float]
123- ==== Disable `norms` and `doc_values` on sparse fields
124-
125- If none of the above recommendations apply in your case, you might want to
126- check whether you actually need `norms` and `doc_values` on your sparse fields.
127- `norms` can be disabled if producing scores is not necessary on a field, this is
128- typically true for fields that are only used for filtering. `doc_values` can be
129- disabled on fields that are neither used for sorting nor for aggregations.
130- Beware that this decision should not be made lightly since these parameters
131- cannot be changed on a live index, so you would have to reindex if you realize
132- that you need `norms` or `doc_values`.
133-
0 commit comments