Right Sizing Policy Parameters

These policy settings impact the output of your Right Sizing Recommendations reports.

You can modify the Default policy parameters to visualize "what if" scenarios. For example, you can compare the results of increasing or decreasing the recommendation aggressiveness, or restricting memory so it cannot decrease versus making recommendations based on historical utilization, etc.

If parameter values are set that cannot be met by some entities, those entities display in the Recommendations table with gray highlighting. This indicates that no active recommendations are possible for the highlighted entities. An example of a constraint that would result in no active recommendations is if you set minimum memory to 50 GB, but the required instance type is set to T3a.small, which does not have 50GB of memory.

Some additional constraints, that cannot be modified, are implemented by default on all right-sizing calculations. For example, right-sizing recommendations assume that the region configured by the user should not change, so recommendations will be kept within the existing assigned regions. Neither does right-sizing address preinstalled licensing, such as a pay-as-you-go Windows license, so license considerations are not included in recommendations.

Right Sizing Constraints Settings

Define minimum requirements based on numerical settings or historical utilization.

. Basic Right Sizing Constraints

Data Aggregation Method

Provides context for how historical utilization data is applied to recommendations.

  • Aggressiveness:

    • 95th Percentile: Considers the behavioral norm demonstrated in utilization data (without edge cases).

    • Min: Considers the behavioral minimum demonstrated in all utilization data.

    • Max: Considers the behavioral maximum demonstrated in all utilization data, regardless of rarity.

CPU

The following constraints can be applied to more strictly define your minimal resource requirements for CPUs.

  • No Constraint: Returns all possible recommendations for the given resource.

  • CPUs cannot decrease : Returns recommendations where maximum resource utilization is at or below the input amount.

  • Minimum CPUs: Returns recommendations where minimum resource utilization is at or above the input amount.

  • Resize based on historical utilization: Returns recommendations based on historical utilization of the given resource type.

    • Target CPU utilization: Set a specific utilization target as a percentage of CPU resources. Returns recommendations that closely match the target CPU utilization percentage.

Memory

The following constraints can be applied to more strictly define your minimal resource requirements.

  • No Constraint: Returns all possible recommendations for the given resource.

  • Memory cannot decrease: Returns recommendations where maximum resource utilization is at or below the input amount.

  • Minimum Memory (GB): Returns recommendations where minimum resource utilization is at or above the input amount.

  • Resize based on historical utilization: Returns recommendations based on historical utilization of the given resource type.

    • Default memory utilization when diagnostic monitoring not enabled: Specify memory utilization (by percent) when it does not exist because guest OS diagnostic metrics have not been enabled on the cloud provider. Returns recommendations based on an assumption (as a percentage) of memory utilization.

      If diagnostic metrics are not enabled and this option is not enabled, then no memory utilization data can be provided.

      60% is the recommended setting unless a more accurate percentage is determined.

    • Target memory utilization: Set a specific utilization target as a percentage of memory resources. Returns recommendations that closely match the target memory utilization percentage.

. Instance Type Constraints

You can specify target instance types (e.g., a series) that can be proposed by the Right Sizing Report. This section displays instance types based on the report's data source type for Azure and AWS.

  • No Constraint, select from all available instance types: Returns all possible recommendations for the given resource.

  • Include selected instance types only.

  • Exclude specific instance types.

. Other Resource Constraints
  • Disk I/O

    • No Constraint: Returns all possible recommendations for the given resource.

    • Peak IOPS cannot decrease: Returns recommendations where IOPs are at or above existing level.

    • Minimum peak IOPS: Returns recommendations where IOPs does not go below the peak level.

  • Network I/O

    • No Constraint: Returns all possible recommendations for the given resource.

    • Network performance cannot decrease: Returns recommendations where network performance is at or above existing level.

    • Minimum peak network performance (Gbps): Returns recommendations where network performance does not go below the input peak level.

  • EC2 Entities: Tenancy

    • No Constraint: Returns all possible recommendations for the given attribute.

    • Shared: Returns only shared tenancy options.

    • Dedicated: Returns only dedicated tenancy options.

    • Host: Returns only host tenancy options.

  • EC2 Entities: Term Type

    • No Constraint: Returns all possible term recommendations for the given attribute.

    • Reserved: Returns only recommendations with reservation terms.

      • Reservation Offering Class: Options include Any, Standard, or Convertible.

      • Reservation Contract Length: Options include Any, 1 year, or 3 years.

      • Reservation Purchase Option: Options include Any, No Upfront, All Upfront, Partial Upfront.

      • MS SQL Server: Options include Any, NA, SQL Ent, SQL Std, SQL Web.

    • OnDemand: Returns only recommendations with On-Demand terms.

  • EC2 Entities: Graviton CPU

    • No Constraint: Returns all possible recommendations for the given attribute.

    • Dynamic: Returns recommendations either including or excluding Graviton CPUs.

      • Only Graviton CPU

      • Exclude Graviton CPU

    • No Change: