[
https://issues.apache.org/jira/browse/LUCENE-9378?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17196567#comment-17196567
]
Julie Tibshirani commented on LUCENE-9378:
------------------------------------------
Adding another datapoint in case it’s helpful: Elasticsearch uses binary doc
values to store high-dimensional float vectors (the `dense_vector` type). The
vectors are typically used to score documents. This is a very similar use case
to the one [~alexklibisz] describes.
In a best case scenario where we index random 100-dimensional float vectors (so
they don’t compress well), and scan over all documents (so we don’t decompress
data that’s not needed), we see a QPS drop of ~10%:
{{Algorithm QPS}}
{{Old version 66.989}}
{{New version 60.906}}
I would guess [~nicot]'s example also falls under the general category of
"storing/ loading numeric feature vectors"? For this type of data, it’s common
to take a specialized approach instead of applying a general compression
algorithm (like using a sparse vector representation, using fewer bytes for
each vector element, etc.)
> Configurable compression for BinaryDocValues
> --------------------------------------------
>
> Key: LUCENE-9378
> URL: https://issues.apache.org/jira/browse/LUCENE-9378
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Viral Gandhi
> Priority: Minor
> Attachments: hotspots-v76x.png, hotspots-v76x.png, hotspots-v76x.png,
> hotspots-v76x.png, hotspots-v76x.png, hotspots-v77x.png, hotspots-v77x.png,
> hotspots-v77x.png, hotspots-v77x.png, image-2020-06-12-22-17-30-339.png,
> image-2020-06-12-22-17-53-961.png, image-2020-06-12-22-18-24-527.png,
> image-2020-06-12-22-18-48-919.png, snapshot-v77x.nps, snapshot-v77x.nps,
> snapshot-v77x.nps, snapshots-v76x.nps, snapshots-v76x.nps, snapshots-v76x.nps
>
> Time Spent: 4h 10m
> Remaining Estimate: 0h
>
> Lucene 8.5.1 includes a change to always [compress
> BinaryDocValues|https://issues.apache.org/jira/browse/LUCENE-9211]. This
> caused (~30%) reduction in our red-line QPS (throughput).
> We think users should be given some way to opt-in for this compression
> feature instead of always being enabled which can have a substantial query
> time cost as we saw during our upgrade. [~mikemccand] suggested one possible
> approach by introducing a *mode* in Lucene80DocValuesFormat (COMPRESSED and
> UNCOMPRESSED) and allowing users to create a custom Codec subclassing the
> default Codec and pick the format they want.
> Idea is similar to Lucene50StoredFieldsFormat which has two modes,
> Mode.BEST_SPEED and Mode.BEST_COMPRESSION.
> Here's related issues for adding benchmark covering BINARY doc values
> query-time performance - [https://github.com/mikemccand/luceneutil/issues/61]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
