Add notes on indexing to kNN search guide by jtibshirani · Pull Request #83188 · elastic/elasticsearch

jtibshirani · 2022-01-27T00:19:08Z

This change adds a new 'indexing considerations' section that explains why index
calls can be slow and how force merge can help search latency.

This change adds a new 'indexing considerations' section that explains why index calls can be slow and how force merge can help search latency.

jtibshirani · 2022-01-27T00:20:45Z

docs/reference/search/search-your-data/knn-search.asciidoc


+[discrete]
+[[knn-indexing-considerations]]
+==== Indexing considerations


I'm very open to suggestions for a better name :)

What is the reason we put this section in knnSearch guide? Isn't a better place for it the indexing section of dense-vector document?
Or we provide these kind of suggestions only in reference guides as this one?

I was wondering this myself and would appreciate @jrodewig's opinion here. I was thinking of this reference as a combined "how to" for kNN search, where we offer guidance about what method to choose, other best practices, etc. I think of the API and field type docs more as specific references that don't give high-level guidance like this.

I'm very open to suggestions for a better name :)

I think Indexing considerations is fine. However, you could do Indexing speed if wanted.

I was thinking of this reference as a combined "how to" for kNN search, where we offer guidance about what method to choose, other best practices, etc. I think of the API and field type docs more as specific references that don't give high-level guidance like this.

Yep. That's pretty much my model of these docs. I'm okay with linking from the dense-vector docs to this section, but I think this content is a little too dense and specific to approximate kNN search for the dense-vector reference docs.

elasticmachine · 2022-01-27T00:21:01Z

Pinging @elastic/es-search (Team:Search)

elasticmachine · 2022-01-27T00:21:01Z

Pinging @elastic/es-docs (Team:Docs)

docs/reference/search/search-your-data/knn-search.asciidoc

mayya-sharipova

@jtibshirani Thanks Julie, the changes LGTM. I am just not sure what's the best place for this next section.

docs/reference/search/search-your-data/knn-search.asciidoc

mayya-sharipova · 2022-01-27T15:47:59Z

docs/reference/search/search-your-data/knn-search.asciidoc


+[discrete]
+[[knn-indexing-considerations]]
+==== Indexing considerations


What is the reason we put this section in knnSearch guide? Isn't a better place for it the indexing section of dense-vector document?
Or we provide these kind of suggestions only in reference guides as this one?

jrodewig

LGTM overall with @mayya-sharipova's edits.

I left some suggestions that swap the ANN reference to HNSW graphs, but I'd like for you to take a look before accepting it. The other suggestions are non-blocking. Feel free to ignore those if wanted.

Thanks @jtibshirani!

jrodewig · 2022-01-27T21:53:46Z

docs/reference/search/search-your-data/knn-search.asciidoc


+[discrete]
+[[knn-indexing-considerations]]
+==== Indexing considerations


I'm very open to suggestions for a better name :)

I think Indexing considerations is fine. However, you could do Indexing speed if wanted.

I was thinking of this reference as a combined "how to" for kNN search, where we offer guidance about what method to choose, other best practices, etc. I think of the API and field type docs more as specific references that don't give high-level guidance like this.

Yep. That's pretty much my model of these docs. I'm okay with linking from the dense-vector docs to this section, but I think this content is a little too dense and specific to approximate kNN search for the dense-vector reference docs.

docs/reference/search/search-your-data/knn-search.asciidoc

jrodewig · 2022-01-27T22:09:11Z

docs/reference/search/search-your-data/knn-search.asciidoc

+Indexing vectors for approximate kNN search can take substantial time because
+of how expensive it is to build the ANN index structures. You may need to
+increase the client request timeout for index and bulk requests.


Instead of ANN, we may want to mention HNSW graphs here. We talk about them in the next paragraph without much of an introduction.

I took a stab at a suggestion, but feel free to edit or ignore if wanted.

Suggested change

Indexing vectors for approximate kNN search can take substantial time because

of how expensive it is to build the ANN index structures. You may need to

increase the client request timeout for index and bulk requests.

{es} shards are composed of segments, which are internal storage elements in the

index. For approximate kNN search, {es} stores the dense vector values of each

segment as an https://arxiv.org/abs/1603.09320[HNSW graph]. Indexing vectors for

approximate kNN search can take substantial time because of how expensive it is

to build these graphs. You may need to increase the client request timeout for

index and bulk requests.

This reorganization works nicely, I'll adopt it.

docs/reference/search/search-your-data/knn-search.asciidoc

jtibshirani · 2022-01-28T18:23:26Z

Thanks for the helpful comments!

This change adds a new 'indexing considerations' section that explains why index calls can be slow and how force merge can help search latency.

* upstream/master: (100 commits) Avoid duplicate _type fields in v7 compat layer (elastic#83239) Bump bundled JDK to 17.0.2+8 (elastic#83243) [DOCS] Correct header syntax (elastic#83275) Add unit tests for indices.recovery.max_bytes_per_sec default values (elastic#83261) [DOCS] Add note that write indices are not replicated (elastic#82997) Add notes on indexing to kNN search guide (elastic#83188) Fix get-snapshot-api :docs:integTest (elastic#83273) FilterPathBasedFilter support match fieldname with dot (elastic#83178) Fix compilation issues in example-plugins (elastic#83258) fix ClusterStateListener javadoc (elastic#83246) Speed up Building Indices Lookup in Metadata (elastic#83241) Mute whole suite for elastic#82502 (elastic#83252) Make PeerFinder log messages happier (elastic#83222) [Docs] Add supported _terms_enum field types (elastic#83244) Add an aggregator for IPv4 and IPv6 subnets (elastic#82410) [CI] Fix 70_time_series/default sort yaml test failures (elastic#83217) Update test-failure Issue Template to include "needs:triage" label elastic#83226 Add an index->step cache to the PolicyStepsRegistry (elastic#82316) Improve support for joda datetime to java datetime transition in Painless (elastic#83099) Fix joda migration for week based methods in Painless (elastic#83232) ... # Conflicts: # x-pack/plugin/rollup/src/main/java/org/elasticsearch/xpack/rollup/v2/TransportRollupAction.java

Add notes on indexing to kNN search guide

5100f32

This change adds a new 'indexing considerations' section that explains why index calls can be slow and how force merge can help search latency.

elasticsearchmachine added the v8.1.0 label Jan 27, 2022

jtibshirani commented Jan 27, 2022

View reviewed changes

jtibshirani added :Search/Search Search-related issues that do not fall into other categories >docs General docs changes v8.0.0 labels Jan 27, 2022

elasticmachine added Team:Search Meta label for search team Team:Docs Meta label for docs team labels Jan 27, 2022

jrodewig self-requested a review January 27, 2022 02:10

mayya-sharipova reviewed Jan 27, 2022

View reviewed changes

docs/reference/search/search-your-data/knn-search.asciidoc Outdated Show resolved Hide resolved

mayya-sharipova approved these changes Jan 27, 2022

View reviewed changes

jrodewig approved these changes Jan 27, 2022

View reviewed changes

jtibshirani added 2 commits January 28, 2022 10:02

Merge remote-tracking branch 'upstream/master' into knn-docs

a50ba35

Address review feedback

58f9940

jtibshirani merged commit e7ba03e into elastic:master Jan 28, 2022

jtibshirani deleted the knn-docs branch January 28, 2022 18:23

jtibshirani added :Search Relevance/Vectors Vector search and removed :Search/Search Search-related issues that do not fall into other categories labels Jul 21, 2022

-Indexing vectors for approximate kNN search can take substantial time because
-of how expensive it is to build the ANN index structures. You may need to
-increase the client request timeout for index and bulk requests.
+{es} shards are composed of segments, which are internal storage elements in the
+index. For approximate kNN search, {es} stores the dense vector values of each
+segment as an https://arxiv.org/abs/1603.09320[HNSW graph]. Indexing vectors for
+approximate kNN search can take substantial time because of how expensive it is
+to build these graphs. You may need to increase the client request timeout for
+index and bulk requests.

Conversation

jtibshirani commented Jan 27, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

elasticmachine commented Jan 27, 2022

Uh oh!

elasticmachine commented Jan 27, 2022

Uh oh!

Uh oh!

mayya-sharipova left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jrodewig left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jtibshirani commented Jan 28, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants