Page tree

Unsupported content

   

This version of the documentation is no longer supported. However, the documentation is available for your convenience. You will not be able to leave comments.

Skip to end of metadata
Go to start of metadata

This topic explains how to configure additional placement options and Affinity settings for Compute and Storage resources, for use in Capacity-Aware Placement Advice (CAPA).

Setting placement options and Affinity for Compute and Storage Resources

To configure placement options and Affinity settings for Compute and Storage resources, perform the following task:

  1. Navigate to Administration > ETL & Systems Tasks > System Tasks and select the BMC - Cloud Lifecycle Management - Placement Advice Preparation task from the System tasks table.
  2. In the task details page, click  under Run configurations to edit the active run configuration.
  3. In the Edit run configuration Default page, click the Placement advice algorithm tab and select the appropriate placement and Affinity settings as described in the following table:

    Settings marked with (tick) are available only in Advanced mode.
    Setting ↓OptionDescription
    Analysis Periodhour, day, week

    The time duration for which data is analyzed to compute capacity metrics, which CAPA uses to make decisions.

    The time duration can be set as number of hours, days, or weeks dating back from the current time or the last completed time depending on whether the include current time check box is selected or not. The check box label changes depending on the option you select in the list box. For example, if you select hour in the list, the check box name changes to include current hour.

    The default setting for this option is 24 hours back, including current time.

    This option is similar to the one available for the time forecasting scenario.

    (tick)Aggregation FunctionAverage, Min, Max

    This statistic is used to compute the day profile of the capacity metrics, across days or weeks of the selected analysis period.

    You can set this value as Average (default), Min, or Max.

    If the default analysis period is used, this option has no effect.

    Scenario PPA Time ValidityNone

    The number of minutes (default: 60) after which Pending Provisioning Action (PPA) is no longer valid.

    This value must be equal to the vCenter Service Extractor Saver Period.

    For more information, see Pending Provisioning Action (PPA) simulations.

    (tick)Scenario Refresh PeriodNone

    The frequency with which the CAPA scenarios are forced to refresh, to load utilization data, that is fed by the performance ETLs.

    By default, this option is set to a value that is half of the value set for the Scenario PPA Time validity (30 minutes).
    Compute resources
    Affinity for compute resources

    Prefer SOI

    Try to use the same target for all VMs of an SOI, if possible.
    Strict SOIUse one target for all VMs of an SOI, or fail provisioning.
    No affinityUse the target that has the most available space for each VM.
    For compute resources, fail placement   When insufficient capacity is detected or anticipated using simulation of pending provisioning actionsFail the placement for Compute Resources if Insufficient Capacity is detected while considering Simulations provided by Pending Provisioning Actions.
    When insufficient capacity is detected (recommended)Fail the placement for Compute Resources if Insufficient Capacity is detected without considering Simulations provided by Pending Provisioning Actions.
    When insufficient capacity is detected, but never for compute pools of Virtual Resource PoolsFail the placement for Compute Resources if Insufficient Capacity is detected, without considering Simulations provided by Pending Provisioning Actions in all calls, except those that include compute pools of Virtual Resource Pools.
    NeverNever fail the placement for Compute Resources.
    Capacity estimation for compute pool of Virtual Resource PoolsOptimistic: Use unreserved capacity for the parent Cluster (Default and recommended option)The metric of a VRP for a resource is the Reserved VRP plus a fraction of the unreserved resource on the cluster whose resource the VRP is partitioning. For more information, see optimistic computation.
    Pessimistic: Use a portion of the unreserved parent Cluster capacity given by Shares percentage assigned to the Virtual Resource PoolUse the entire unreserved available capacity of the cluster, instead of the fraction (depending on the shares) allocated to the VRPs contained in the Compute Pool. For more information, see pessimistic computation.
    Storage Resources
    Affinity for storage resourcesPrefer VMTry to use the same datastore for all vDisks of a VM, if possible.
    Strict VMUse one datastore for all vDisks of a VM, or fail provisioning.
    Prefer SOITry to use same datastore for all vDisks of all VMs of the SOI, if possible.
    Strict SOIUse one datastore for all vDisks of all VMs of an SOI, or fail provisioning.
    No affinityUse the datastore that has the most available space for each vdisk.
    For storage resources, fail placement



    When insufficient capacity is detected or anticipated using simulation of pending provisioning actions (recommended)Fail the placement for Storage Resources if Insufficient Capacity is detected while considering Simulations provided by Pending Provisioning Actions.
    When insufficient capacity is detectedFail the placement for Storage Resources if Insufficient Capacity is detected without considering Simulations provided by Pending Provisioning Actions.
    NeverNever fail the placement for Storage Resources.
  4. Click Save.

The following image shows an example of the Placement Advice Algorithm tab in the Edit run configuration page.

Placement Advice Algorithm tab

Capacity estimation for compute pool of Virtual Resource Pools

For CLM Compute Pools, capacity metrics in BMC TrueSight Capacity Optimization are estimated through an aggregation of metrics of contained entities.

Specifically, in case of a Compute Pool of Virtual Resource Pools (VRP), the Total clock frequency (CPU_TOTAL_MHZ) and Real memory (TOTAL_REAL_MEM), which is referred to as SIZE in further computations, can be calculated using the following options:

  • Optimistic (default)
  • Pessimistic

Recommendation

Use the (default) Optimistic option to calculate the Compute Pool of VRPs.

The following table describes these options along with their computations in detail.

OptionComputations

Optimistic: Use unreserved capacity for the parent Cluster

(Default and recommended option)

SIZE(Compute Pool) =
Sum[i:Contained VRPs](COMPUTED_LIMIT_PESS(i))

COMPUTED_LIMIT_PESS: Determines how a resource available to a given VRP is expressed, when cluster resources are in maximum contention.

The metric of a VRP for a resource is the total of the Reserved VRP and a fraction of the unreserved resource on the cluster whose resource the VRP is partitioning. It is calculated as:

COMPUTED_LIMIT_PESS(VRP) =
RESERVED (VRP) + (UNRESERVED(Cluster) *
SHARE(VRP))/SUM[i : VRPs in Cluster] (SHARE(i)))
Pessimistic: Use a portion of the unreserved parent Cluster capacity given by Shares percentage assigned to the VRP
SIZE (Compute Pool) =
UNRESERVED(Cluster) + Sum[i : Contained VRPs] (RESERVED(i))

This computation uses the entire unreserved available capacity of the cluster, instead of the fraction (depending on the shares) allocated to the VRPs contained in the Compute Pool.

Related topic

Affinity for Capacity-Aware Placement Advice overview