Dynamic Resource Optimizer

Edit online

IBM® PowerVC Dynamic Resource Optimizer (DRO) continuously monitors the resource utilization of physical hosts and virtual machines in a cloud environment. When usage imbalances are detected, the PowerVC DRO optimizes the cloud by making adjustments for workloads whose characteristics change over time. Capacity on Demand (CoD) from IBM Power Systems is a powerful tool that activates unused cores or mobile memory for a host, giving it more computing power or memory instantly. On eligible systems, the PowerVC DRO works with CoD to relieve overburdened systems. For systems that are not Enterprise Pool members, DRO migrates virtual machines.

What can DRO do for you?

DRO's automated processes save your time by automating optimization in your environment and by giving you an easy way to view those changes. DRO allows you to do the following:

  • Set up DRO to monitor your cloud's utilization and address either CPU or memory resource imbalances.
  • Specify different acceptable levels of usage for each host group.
  • Run DRO in Advise mode to see where usage problems are occurring without letting DRO act on your environment. You can change between Active and Advise modes at any time.
  • Take advantage of Power Systems Enterprise Pool CoD licenses by assigning cores or mobile memory to overburdened hosts.
  • Specify which action is taken: a CoD operation (activation of cores or mobile memory, as appropriate), virtual machine migration, or both.
  • Exclude specific hosts or virtual machines to control which resources are optimized and ensure that mission-critical virtual machines are not migrated.
  • Easily see which hosts are in violation of the optimization policy so you know which hosts are subject to DRO operations.
  • Quickly determine what optimization events occurred.

Example of mobile core activation

The following example is for mobile core activation, but mobile memory is assigned in a similar fashion.

When a host's CPU utilization is imbalanced in relation to a user-defined threshold and is a member of an Enterprise Pool, DRO tries to use CoD to optimize host usage. This is much faster than migrating virtual machines and does not burden the network. The following animation illustrates how the DRO can activate mobile cores as the CPU utilization for a host becomes overburdened:
This figure shows how DRO can activate mobile cores as workload demands change
When a host becomes over-utilized and is not a member of an Enterprise Pool, virtual machines from the violated host are migrated to less-burdened hosts. The following animation shows three hosts in a host group that have DRO enabled. The hosts are not members of an Enterprise Pool. When Host 1 becomes overburdened, the virtual machines are migrated to other hosts in the host group:
This figure shows how DRO can migrate an over-utilized virtual machine

What DRO monitors

DRO can monitor one of these resources. You can change which resource is monitored at any time:

CPU
DRO monitors the host's CPU usage. This is available for all hosts. DRO can migrate virtual machines or activate mobile cores to correct imbalances.
Memory
DRO monitors the host's available memory. DRO can migrate virtual machines or activate mobile memory to correct imbalances. This is only available for PowerVM® NovaLink managed hosts.

DRO actions

The actions that are taken by DRO depend on the values that you select under Resource to monitor. You can select any combination of these actions. In each optimization cycle, DRO performs up to one action on each host:

Migrate virtual machines
DRO can always migrate virtual machines to correct a resource imbalance. At most one virtual machine is migrated per host, up to the maximum allowed concurrent migrations for the host group.
Activate mobile cores or memory
The host must be a member of an Enterprise Pool to be a candidate for CoD operations, that is, mobile core or mobile memory activation. If you configure DRO for both actions, the CoD operation is always the preferred action. These actions are instantaneous and do not use any network resources. In comparison, migrating a virtual machine can take several minutes and uses your network to transfer the data.
Note: DRO ignores hosts that are in maintenance mode. Such hosts are not monitored and are not targets for migrations.

How DRO becomes activated

Each host group has an optimization policy that specifies what it means for a host to be overburdened. These values define the optimization policy: Resource to monitor, Utilization threshold, Run interval, and Stabilization. DRO acts only after the Utilization threshold for the Resource to monitor is crossed the number of times that are specified in Stabilization.

For example, assume that you configured DRO to monitor the CPU utilization and set these values:
  • Resource to monitor: CPU
  • Utilization threshold = 70%
  • Run interval = 5 minutes
  • Stabilization = 3 times
Then DRO checks the utilization on every host in the host group every five minutes. If any host has a utilization greater than 70%, its threshold is exceeded. If the host's utilization exceeds the utilization threshold three times in a row, the appropriate DRO action is triggered. This action could be activating a free core on the host, activating mobile memory, or migrating a virtual machine off the host, depending on the operations that you selected.