L12.0 Metrics concepts15%

How a metric is storedWhat a metric actually is

Charts, analytics and detectors all run on metric time series. Get this layer exact and the rest of the exam gets easier.

Everything else on the exam is built on this one idea. Nail it and the rest feels obvious.

LAB 01

Take a datapoint apartOne reading, five parts

Picture a postcard your server mails every 10 seconds. It says what it measured, what kind of measurement, the number, a return address saying who sent it, and sometimes a date. That postcard is a datapoint.

Try it in your Observability Cloud org
  1. Open Metric Finder and search for a CPU metric your org already sends.
  2. Open it in a chart and look at the dimensions listed for each series.
  3. Name which part of a datapoint each thing on screen is.
LAB 02

Build metric time seriesEvery sender gets its own diary

An MTS is every datapoint that shares a metric name, metric type and the exact same set of dimensions. Change the dimensions and watch the series count. Then add a dimension key, or attach a custom property, and compare.

Postcards with the same "what" and the exact same return address go in one diary. That diary is a metric time series (MTS). Add even one line to the return address and the mailroom starts a new diary. A sticky note on the cover is different: add or change it any time and it never creates a new diary. That's a custom property.

Metric
cpu.utilization · gauge
Dimension · host
Dimension · env
0active MTSdiaries getting pages
0inactive, kept 13 monthsclosed, on the shelf
Try it in your Observability Cloud org
  1. Chart a metric with no filter and count the lines.
  2. Filter by one dimension value, then add a second filter, and count again.
  3. Explain the change in plain words: each line is one MTS.
LAB 03

Watch a rollup reshape the chartSquishing lots of readings into one

Datapoints arrive every 10 seconds. At a coarser chart resolution, a rollup summarizes each window into one value. The metric type picks the default (★). The cumulative counter restarts near minute 9: Delta never goes negative.

A chart can't fit 90 dots in a small space, so it groups them into buckets, like one per minute, and boils each bucket down to one number. That's a rollup. How you boil it down depends on the meter: a speedometer gets averaged, a tally clicker gets added up, an odometer gets finish minus start. The star marks the default rollup the exam expects.

Metric type
Resolution
Rollup
Raw points90
On chart15
You're seeingAverage of each minute
Try it in your Observability Cloud org
  1. Open a chart on a busy metric and set the time range to the last 15 minutes.
  2. Switch to the last 7 days and watch the line smooth out as each point covers a longer window.
  3. Open the plot's settings and switch the rollup, then compare the shape.
REFERENCE

Metadata and metric typesThe four meters and three labels

DimensionsCustom propertiesTags
AddedWith the data, at ingestAfter ingestAfter ingest
Editable laterNoYesYes
Defines the MTSYesNoNo
Values per keyOneOneMany, in sf_tags
Limits36 per MTS · key 128 · value 256 chars··
Metric typeRepresentsDefault rollupExample
GaugeA value at a point in timeAverageCPU %, memory used
CounterOccurrences in an intervalSumRequests in the last 10 s
Cumulative counterRunning total since startDeltaTotal bytes sent
HistogramA distribution with percentilesHistogramRequest latency

Resolution is 10 seconds by default; slower senders keep their native resolution; sf_hires=1 gives 1-second data. Rollups run after data leaves storage and before analytic functions. Inactive MTS are kept 13 months.

metric_name→Metric name
_value→Value
Event dimensions→Dimensions, which define the MTS
mstats avg() span=1m→Rollup at chart resolution

Bottom row: Splunk Enterprise on the left, Observability Cloud on the right.

Speedometer

What's true right now: 42% CPU. To sum up a minute, take the average.

Gauge → Average

Tally clicker

Counts clicks, then resets every 10 seconds. To sum up a minute, add them all up.

Counter → Sum

Car odometer

Only goes up. You care how far you drove this trip. A swapped odometer reading 120 means you drove 120, never a negative number.

Cumulative counter → Delta

Sorted pile

Keeps every measurement in order so you can ask how slow the slowest 1% were.

Histogram

Return address, in pen

Sent with every postcard; can't be changed later. Change it and you get a new diary. Max 36 lines, or the postcard is thrown away.

Dimension

Sticky note on the cover

Added later in the app. Edit any time. One value per note. Never makes a new diary.

Custom property

Colored stickers

Added later too, and a diary can wear many. A note on a sticker color applies to every diary wearing it.

Tag · sf_tags
Postcard→Datapoint
Diary→MTS
How many diaries→Cardinality
Boiling a bucket down→Rollup
Postcard every 10 s→10-second resolution
Closed diary on the shelf→Inactive MTS · 13 months
Practice · lesson 1

Check yourself

Questions are worded like the real exam. Press 1 to 4 to answer and Enter to continue.

Questions are worded like the real exam. Read the explanation after each one, even when you get it right.