Creating an SLO

To monitor a service, you need at least one service-level objective (SLO). The SLOs encapsulate your performance goals for the service. Every SLO is based on a performance metric, called a service-level indicator (SLI). For backgound information on SLIs and SLOs, see Concepts in service monitoring.

You can create up to 500 SLOs for a service.

Overview

To define an SLO, navigate to the Create a Service Level Objective (SLO) pane by doing one of the following:

  • For a new service, click the Create an SLO button that is displayed after your new service is created.

  • For an existing service:

    • Click the name of the service in the Services Overview dashboard.
    • Click Create a SLO in the Current status pane on the service's dashboard.

The SLO-creation pane leads you through the steps to create an SLO. The remainder of this section describes each of the following steps in the SLO-creation process:

  1. Set the SLI.
  2. Define SLI details.
  3. Set the SLO.
  4. Review and save the SLO.

To advance to the next step, click Next. To return to the previous step, click Back. To exit the SLO-creation process, click Close.

Setting your SLI

The Set your SLI pane has the following sub-panes:

  • Service details, which reports identifying information about your service. This is the same as the Service details pane on the dashboard for the service.

  • Choose a metric, where you choose a metric for the performance you want to monitor.

  • Request-based or windows-based?, where you choose how the metric is to be evaluated.

The following screenshot shows the SLI pane:

Use the **Set your SLI** pane to choose a performance metric

For more information about metrics used in SLIs and the evaluation methods, see the conceptual topic Service-level indicators.

Choosing a metric

The SLI metric specifies the type of performance you want to measure. In the SLI, you build a ratio from the metric to measure good performance over time. You have the following options for SLIs:

  • Availability, which measures how available your service is to users.
  • Latency, which measure how responsive your service is to users.
  • Other, which lets you indicate that you want to use a specific metric. You specify the metric and describe how to build the SLI on the Set SLI details pane.

The valid choices depend on the type of service you are configuring:

  • For services on Anthos Service Mesh, Istio on Google Kubernetes Engine, and App Engine, you can chose any of the options. The availability and latency metrics are already known for these services, or you choose Other to use a custom SLI.

  • For custom services, the only choice is Other. Meaningful availability or latancy metrics aren't known in advance for custom services, so you must define a custom SLI.

Choosing the evaluation method

After you select the metric for your SLI, you specify how the metric should be evaluated.

  • Request-based evaluation measures the number of requests that meet the evaluation criterion against the total number of requests in a given period.

  • Windows-based evaluation measure the number of evaluation periods that meet a goodness criterion against the total number of evalution periods.

For both evaluation methods, you specify the evaluation criteria on the Set SLI details page.

For more information on these evaluation types, see Compliance in request- and windows-based SLOs.

Setting SLI details

The contents of the Define SLI details pane depends on the metric and evaluation method you chose in the previous step.

If you chose the availability metric and request-based evaluation, there are no other details needed.

Windows-based evaluation

If you selected window-based evalution, you set the additional criteria for the window on this pane: a goodness criterion and a duration.

Set the SLI window by choosing a goodness criterion and evaluation period.

The goodness criterion indicates the percentage of windows that must evaluate to “good” over the compliance period. The duration specifies the length of the window.

Latency metric

If you chose the latency metric, you specify the threshold value that determines acceptable performance on this pane:

Set the latency threshold for the SLI.

Anything above the latency threshold is considered “bad” performance in evaluating the SLI.

Custom SLI

If you selected Other as the SLI metric, you specify the metric you want to use on this pane. You can select a metric by typing in the Performance Metric field or select one from the list.

The metrics in the list are divided into two types:

  • Distribution-cut indicators
  • Time-series ratio indicators

The following screenshot shows a partial list:

Metrics in the menu are classified by indicator type.

If you select a distribution-cut indicator, you configure the SLI by providing a range—above, below, or between—and a filter to specify the monitored resource and any labels you want to include. The configuration pane looks like the following:

Set a range and filter for a distribution-cut indicator.

If you select a time-series ratio indicator, you configure the ratio by building numerator and denominator filters to classify the metric data, typically by selecting the values of labels in the metric or resource type. The configuration pane looks like the following:

Set numerator and denominator filters for a time-series ratio.

For more information about these SLI types, see the Monitoring API reference pages for DistributionCut and TimeSeriesRatio.

Preview chart

After you have configured the SLI, the Define SLI details pane includes a preview chart to show you how the historical performance of this service is measured by the SLI. For example:

The completed SLI shows a chart based on historical data.

If you have just created or deployed a service, there may not be any data yet. You can still create the SLI, but you won't get the historical perspective.

Setting your SLO

The Set your SLO pane has the following regions:

  • Compliance period, where you set the time period over which you want to evaluate the SLI.

  • Performance goal, where you specify the threshold for performance over the compliance period.

  • Preview, which displays a chart that shows the performance-goal threshold and graph that shows the results of evaluating the SLI over the compliance period.

Set the SLO by choosing a compliance period and a performance goal.

Compliance period

There are two types of compliance period, which you select from the menu:

  • Calendar period
  • Rolling window

A calendar period measures compliance over a fixed period of time, the period length. When the period ends, the error budget is reset and a new compliance period starts.

A rolling window is a sliding period. It also has a length, but the compliance is computed over the last n days. When a new day starts, the compliance and remaining error budget are recomputed over the previous n days.

For more on calendar and rolling-window compliance periods, see Compliance periods.

Preview chart

After you have configured the SLO, the Set your SLO pane includes a preview chart to show you how the historical performance of this service is measured by the SLO. For example:

The completed SLO shows a chart based on historical data.

If you have just created or deployed a service, there may not be any data yet. You can still create the SLO, but you won't get the historical perspective.

Saving your SLO

The Review and save pane has a single field, a display name for the SLO. The field has a default value based on the selections you made while defining the SLO, but you can change it to make the display name more descriptive.

The following screenshot shows the field with a default name:

Monitoring generates a default name for your SLO.

When you are satisfied with the display name, click Submit.

After you create an SLO, you can create an alerting policy to monitor it. For more information, see Alerting on your burn rate and Creating an alerting policy.