[[release-highlights-7.6.0]]
== 7.6.0 release highlights
++++
<titleabbrev>7.6.0</titleabbrev>
++++

//NOTE: The notable-highlights tagged regions are re-used in the
//Installation and Upgrade Guide

// tag::notable-highlights[]
[float]
==== New histogram field type

A new {ref}/histogram.html[histogram field type] has been added. The new `histogram` field accepts
pre-aggregated histograms which can later be used directly in the
{ref}/search-aggregations-metrics-percentile-aggregation.html[percentiles] and
{ref}/search-aggregations-metrics-percentile-rank-aggregation.html[percentile_ranks] aggregations.
This allows users to pre-aggregate histogram data locally and only send the final
data structure, saving storage and network bandwidth while retaining the ability to
aggregate it like any other data.

// end::notable-highlights[]

// tag::notable-highlights[]
[float]
==== Optimized sorting on long field types

Sorting on a {ref}/number.html[`long`] field now internally rewrites into a Lucene `DistanceFeatureQuery`.
This lets {es} skip non-competitive hits, which often improves query speed.
In benchmarking tests, this sped up sorts on `long` fields by 10x.

// end::notable-highlights[]

// tag::notable-highlights[]
[float]
==== Simplifying and operationalizing machine learning

With the release of 7.6 the {stack} delivers an end-to-end {ml} pipeline
providing the path from raw data to building, testing, and deploying {ml} models
in production. Up to this point {ml} in the {stack} had primarily focused on 
unsupervised techniques by using sophisticated pattern recognition that builds
time series models used for {anomaly-detect}. With the new {dfanalytics}, you
can now use labelled data to train and test your own models, store those models
as {es} indices, and use {ml-docs}/ml-inference.html[inference] to add predicted 
values to the indices based on your trained models.

One packaged model that we are releasing in 7.6 is
{ml-docs}/ml-lang-ident.html[{lang-ident}]. If you have documents or sources
that come in a variety of languages, {lang-ident} can be used to determine the 
language of text so you can improve the overall search relevance.
{lang-ident-cap} is a trained model that can provide a prediction of the
language of any text field.
// end::notable-highlights[]

// tag::notable-highlights[]
[float]
==== {ccs-cap} in {transforms}

{ref}/transforms.html[{transforms-cap}] can now use {ccs} (CCS) for the source
index. Now you can have separate clusters (for example, project clusters) build 
entity-centric or feature indices against a primary cluster.

// end::notable-highlights[]

[float]
=== Learn more

Get more details on these features in the
https://www.elastic.co/blog/elasticsearch-7-6-0-released[{es} 7.6 release blog].
For a complete list of enhancements and other changes, check out the
<<release-notes-7.6.0,{es} 7.6 release notes>>.