Auto-scale Insights

View Auto-scale Insights

The auto-scale insights feature presents a graphical representation of the historic usage and predicted auto-scale requirements for your host pool. It enables you to discover anomalies and to view capacity vs demand in hourly increments.

To view auto-scale insights for a dynamic host pool:

1. Locate the dynamic host pool you want to work with.

2. From the action menu, select Auto-scale > Configure.

   

3. Type the desired Host Session Density.

   Note: Session density is especially important when the auto-scale is configured using non-session criteria. For example, if auto-scale is configured using CPU Usage instead of Available Sessions, auto-scale insights needs to be told the session density. This should be based on the average density you observe in the environment.

4. Optionally, select the auto-scale insights pop-up icon , and then select Load Recommendation to generate the AI auto-scale intelligent recommendations.

   Note: See Auto-scale Insights - Intelligent Recommendations for details about enabling this feature and its workflow.

   This is a sample recommendation:

   

Understanding the auto-scale insights:

The following are important auto-scale insights features.

• Day: At the top of the window, select the desired day to display. This provides usage information for the previous 7 days.

• Hover: You can hover over any part of the graph to see its details. For example:

  

• Statuses:

  • High Over-Provision: This bar shows the large number of available sessions provisioned that were not needed.

  • Over-Provision: This bar shows the moderate number of available sessions provisioned that were not needed.

  • Optimal Provision: This bar shows that the optimal number of sessions were provisioned to meet demand.

  • Under Provision: This bar shows that not enough sessions were provisioned to meet demand.

  • Total Sessions: This bar shows the actual number of total sessions.

  • Excess Users: This bar shows the number of excess users. These are users that attempted to connect to a session, but no sessions were available.

• Bar Type: Solid or Hashed

  

  • Solid: A solid bar represents the minimum available sessions during the recorded time period. The interval for this is 30 minutes.

  • Hashed: A hashed bar represents the maximum available sessions during the recorded time period, The interval for this is 30 minutes.

  Auto-scale Insights Technical Details

  This section provides technical details about how the graph details are calculated.

  Data Collection

  • Duration: Data is collected for the last 7 calendar days based on the host pool's time zone.

  • Exclusion: If sufficient data (2 or more days) is not available, no details are provided.

  Data Sources

  The primary metrics are collected from the following sources:

  • Auto-scale History (Nerdio Manager database)

    • Collection frequency is every 5 minutes by default

    • Metrics:

      • Sessions

      • Host states (drain mode, power state)

  • Host Pool Properties

    • Metric:

      • Maximum sessions per host (if available)

  • Log Analytics (if configured)

    • Metrics:

      • Users with failed login attempts (The count of users that had failed logins due to unavailable free hosts)

      • Users who failed to log in (The count of users that had failed logins due to unavailable free hosts and did not finally connect to a session)

  Data Alignment

  • Available Hosts: Hosts that are running and not in drain mode.

  • Session Density: The lower value between the maximum observed sessions per host or the configured time limit in AVD.

  Due to varying update frequencies from different sources, all metrics are aligned to a specified sample duration, which is 30 minutes by default.

  Interval Calculations

  For each data point within the interval, the following calculations are performed:

  • Sessions Capacity: The total number of sessions the host pool can accommodate, calculated as:

    Sessions Capacity=(Available Hosts×Max Sessions per Host)+Existing Sessions on Other Hosts

  • Total Sessions: The total number of sessions at the given point.

  Aggregated Metrics

  From the above calculations, the following aggregated metrics are produced:

  • Desired Sessions Capacity: The maximum of Total Sessions plus Users who failed to log in.

  • Minimum Sessions Capacity: The minimum of Sessions Capacity.

  • Maximum Sessions Capacity: The maximum of Sessions Capacity.

  Data Analysis - Sample Analysis

  Each data sample is analyzed to determine if any of the following diagnoses apply:

  • Under Provision: This condition is met if:

    • Users with failed logins is greater than 0, irrespective of other metrics.

    • In addition, if the user specifies an exact session density different from the automatically determined one:

      • The under-provisioned status applies when Minimum Sessions Capacity > Desired Sessions Capacity, better accommodating changing density scenarios.

  • Over-Provision: This condition is met if:

    • Minimum Sessions Capacity > Desired Sessions Capacity * factor (1.2 by default)

      In simple terms: Minimum Sessions Capacity is significantly higher than Desired Sessions Capacity.

    • Further checks include:

      • If the difference between Minimum Sessions Capacity and Desired Sessions Capacity is less than 2, it is not considered over-provisioned.

      • If switching off one host would lead to under-provisioning, it is not considered over-provisioned.

  • High Over-Provision This condition is met if:

    • Minimum Sessions Capacity > Desired Sessions Capacity * factor (1.8 by default)

      In simple terms: Minimum Sessions Capacity is much higher than Desired Sessions Capacity.

    • Additional checks include:

      • If the difference between Minimum Sessions Capacity and Desired Sessions Capacity is less than 7, it is not considered a high over-provisioned sample.