OpenSearch/docs/en/ml/aggregations.asciidoc

183 lines
6.3 KiB
Plaintext

[[ml-configuring-aggregation]]
=== Aggregating Data For Faster Performance
By default, {dfeeds} fetch data from {es} using search and scroll requests.
It can be significantly more efficient, however, to aggregate data in {es}
and to configure your jobs to analyze aggregated data.
One of the benefits of aggregating data this way is that {es} automatically
distributes these calculations across your cluster. You can then feed this
aggregated data into {xpackml} instead of raw results, which
reduces the volume of data that must be considered while detecting anomalies.
//TBD: Are "aggregated" and "summarized" equivalent terms? Are customers more
//familiar with one or the other? If so, I'll use one term throughout.
There are some limitations to using aggregations in {dfeeds}, however.
Your aggregation must include a buckets aggregation, which in turn must contain
a date histogram aggregation. This requirement ensures that the aggregated
data is a time series. If you use a terms aggregation and the cardinality of a
term is high, then the aggregation might not be effective and you might want
to just use the default search and scroll behavior.
When you create or update a job, you can include the names of aggregations, for
example:
[source,js]
----------------------------------
PUT _xpack/ml/anomaly_detectors/farequote
{
"analysis_config": {
"bucket_span": "60m",
"detectors": [{
"function":"mean",
"field_name":"responsetime",
"by_field_name":"airline"
}],
"summary_count_field": "doc_count"
},
"data_description": {
"time_field":"time"
}
}
----------------------------------
In this example, the `airline`, `responsetime`, and `time` fields are
aggregations.
NOTE: When the `summary_count_field_name` property is set to a non-null value,
the job expects to receive aggregated input. The property must be set to the
name of the field that contains the count of raw data points that have been
aggregated. It applies to all detectors in the job.
The aggregations are defined in the {dfeed} as follows:
[source,js]
----------------------------------
PUT _xpack/ml/datafeeds/datafeed-farequote
{
"job_id":"farequote",
"indexes": ["farequote"],
"types": ["response"],
"aggregations": {
"buckets": {
"date_histogram": {
"field": "time",
"interval": "360s",
"time_zone": "UTC"
},
"aggregations": {
"time": {
"max": {"field": "time"}
},
"airline": {
"terms": {
"field": "airline",
"size": 100
},
"aggregations": {
"responsetime": {
"avg": {
"field": "responsetime"
}
}
}
}
}
}
}
}
----------------------------------
In this example, the aggregations have names that match the fields that they
operate on. That is to say, the `max` aggregation is named `time` and its
field is also `time`. The same is true for the aggregations with the names
`airline` and `responsetime`. Since you must create the job before you can
create the {dfeed}, synchronizing your aggregation and field names can simplify
these configuration steps.
//TBD: Describe how this would be accomplished in Kibana?
When you define an aggregation in a {dfeed}, it must have the following form:
[source,js]
----------------------------------
"aggregations" : {
"buckets" : {
"date_histogram" : {
"time_zone": "UTC", ...
},
"aggregations": {
"<time_field>": {
"max": {
"field":"<time_field>"
}
}
[,"<first_term>": {
"terms":{...
}
[,"aggregations" : {
[<sub_aggregation>]+
} ]
}]
}
}
}
----------------------------------
You must specify `buckets` as the aggregation name and `date_histogram` as the
aggregation type. For more information, see
{ref}/search-aggregations-bucket-datehistogram-aggregation.html[Date Histogram Aggregation].
NOTE: The `time_zone` parameter in the date histogram aggregation must be set to `UTC`,
which is the default value.
Each histogram bucket has a key, which is the bucket start time. This key cannot
be used for aggregations in {dfeeds}, however, because they need to know the
time of the latest record within a bucket. Otherwise, when you restart a {dfeed},
it continues from the start time of the histogram bucket and possibly fetches
the same data twice. The max aggregation for the time field is therefore
necessary to provide the time of the latest record within a bucket.
You can optionally specify a terms aggregation, which creates buckets for
different values of a field.
IMPORTANT: If you use a terms aggregation, by default it returns buckets for
the top ten terms. Thus if the cardinality of the term is greater than 10, not
all terms are analyzed.
You can change this behavior by setting the `size` parameter. To
determine the cardinality of your data, you can run searches such as:
[source,js]
--------------------------------------------------
GET .../_search {
"aggs": {
"service_cardinality": {
"cardinality": {
"field": "service"
}
}
}
}
--------------------------------------------------
By default, {es} limits the maximum number of terms returned to 10000. For high
cardinality fields, the query might not run. It might return errors related to
circuit breaking exceptions that indicate that the data is too large. In such
cases, do not use aggregations in your {dfeed}. For more
information, see {ref}/search-aggregations-bucket-terms-aggregation.html[Terms Aggregation].
You can also optionally specify multiple sub-aggregations.
The sub-aggregations are aggregated for the buckets that were created by their
parent aggregation. For more information, see
{ref}/search-aggregations.html[Aggregations].
TIP: If your detectors use metric or sum analytical functions, set the
`interval` of the date histogram aggregation to a tenth of the `bucket_span`
that was defined in the job. This suggestion creates finer, more granular time
buckets, which are ideal for this type of analysis. If your detectors use count or rare functions, set
`interval` to the same value as `bucket_span`. For more information about
analytical functions, see <<ml-functions>>.
//TBD: Add more examples from https://github.com/elastic/prelert-legacy/wiki/Configuring-aggregations-on-a-datafeed