pct_sampling Grouping Operator

The percent sampling function, pct_sampling, finds the percentile of a given field. Multiple pct_sampling functions can be included in one query.

The pct_sampling function will return an approximate result for searches that produce large volumes of data.

note

To find the percentile, the function uses a sample of 1,000 messages. This may cause results to be slightly different for different queries when you may expect the results to be the same. 

Syntax

| pct_sampling(<field> [, percentile]) [as <field>] [by <field>]

| pct_sampling(<field> [, percentile, percentile, percentile]) [as <field>] [by <field>]

Rules

• Creates a field with the naming convention: _<field>_pct_<percentile>.
• Separate multiple percent arguments in one query with commas.

Examples

* | parse "data=*" as data
| pct_sampling(data, 95)

Sample log message:

Aug 2 04:06:08 : host=10.1.1.124: local/ssl2 notice mcpd[3772]: filesize=20454: diskutilization=0.4 : 01070638:5: Pool member 172.31.51.22:0 monitor status down.

Example based on sample log message:

file*
| parse "filesize=*" as filesize
| pct_sampling(filesize, 75, 95) by _sourceHost

Running this query creates fields named _filesize_pct_75 and _filesize_pct_95.

A query can also take more multiple percent arguments, such as:

| pct_sampling(q1_delay, 10, 20, 30, 40, 50, 60, 70, 80, 90)

Running this query creates fields named _q1_delay_pct_10, _q1_delay_pct_20, _q1_delay_pct_30, ... to _q1_delay_pct_90.