[Synthetics] Document changing retention on a per-data-stream basis

docs/en/observability/index.asciidoc

@@ -124,6 +124,8 @@ include::inspect-uptime-duration-anomalies.asciidoc[leveloffset=+3]
 
 include::configure-uptime-settings.asciidoc[leveloffset=+2]
 
+include::manage-synthetics-retention.asciidoc[leveloffset=+2]
+
 include::troubleshoot-uptime-mapping-issues.asciidoc[leveloffset=+2]
 
 // User experience

docs/en/observability/manage-synthetics-retention.asciidoc

@@ -0,0 +1,41 @@
+[[manage-synthetics-retention]]
+
+= Managing data retention
+
+[discrete]
+== Overview
+
+Synthetics browser monitors can require large amounts of storage, this document provides information on how synthetics stores data and how to optimize retention
+to control storage utilization. This usually means altering the retention of browser `network` and `screenshot` documents.
+
+All types of checks record 'core' metadata, such as which URL was checked, what the status of the check was, and errors that occurred. This document 
+focuses on browser checks, which tend to use much more storage than lightweight checks. 
+Browser based checks store two additional types of data beyond the core metadata; `network` documents and `screenshot` documents.
+These indices are usually many times larger than the core metadata. Nhe relative sizes of each vary depending on the sites being 
+checked with network data usually being the larger of the two by a significant factor.
+
+`network` documents data consists of detailed metadata around requests for resources required by the pages being checked. An example would be an image, 
+a javascript file, or a font referenced and loaded by the page under test. 
+For each of these requests the URL, timing, headers, and other metadata are stored. While individually this metadata is small, 
+modern websites often request hundreds of additional resources per page load, adding up to a significant amount of storage.
+
+Screenshot data consists of binary image data used to construct a screenshot and metadata with information related to de-duplicating this data. De-duplication 
+efficiency makes screenshot data less burdensome than network data, especially for sites that are mostly visually unchanged across test runs. 
+De-duplication works by splitting each captured screenshot into an 8x8 grid of chunks each stored as a separate document keyed by a hash of its pixels. 
+This means that across checks to a given site only changes to the visual representation of the site require significant additional storage. If a site has not changed
+visually across runs the only additional storage required is the tiny `screenshot_ref` document for that image, which points to the relevant image blocks.
+
+[discrete]
+=== Managing the lifecycles of synthetics data streams
+
+Synthetics data is recorded in Elasticsearch [data streams](https://www.elastic.co/guide/en/elasticsearch/reference/current/data-streams.html), an append-only
+structure in Elasticsearch. Synthetics data streams can be managed via the Elasticsearch API or via [Kibana index management](https://www.elastic.co/guide/en/elasticsearch/reference/current/index-mgmt.html). 
+
+If Synthetics browser data streams are storing data longer than necessary, users can opt to retain `screenshot` and `network_info` datasets for a shorter period than the core metadata.
+To do so, first navigate to [Kibana index management](https://www.elastic.co/guide/en/elasticsearch/reference/current/index-mgmt.html), then filter the list of data streams for
+those containing the term 'synthetics'. In the UI there will be three sorts of data stream present `synthetics-browser-*`, `synthetics-browser.network-*`, and `synthetics-browser.screenshot-*`. From this page you can retrieve the size of each data stream on disk, as well as which [index life cycle](https://www.elastic.co/guide/en/elasticsearch/reference/current/set-up-lifecycle-policy.html) is associated with it. 
+
+To change the retention period of a given data stream simply edit the life cycle policy associated with it after ensuring that policy does not apply to addition data streams whose
+retention you do not want to change. If the data stream you wish to change shares a life cycle policy with another, [create a new ILM policy](https://www.elastic.co/guide/en/elasticsearch/reference/current/set-up-lifecycle-policy.html), then edit the relevant component template's
+index settings. When editing the component template change the `index.lifecycle.name` value to point toward your new ILM policy.  For `screenshot` documents edit the `synthetics-browser.screenshot@package` component template, for `network_info` documents edit the `synthetics-browser.network@package` template. Check that the settings have taken effect by visiting the data stream management
+page and confirming the the attached life cycle policy is correct.