To continue with the cluster enhancements, in this second part we will see some more features, such as improvements to vSphere DRS, scalable shares and DirectPath I/O.
In vSphere 7, the DRS load-balancing algorithm runs every minute. It focuses on the VM DRS score to make recommendations or to migrate VMs:
• Uses VM metrics to calculate the VM DRS score and cluster DRS score
• Identifies VM migrations that improve the VM DRS score
• Uses vSphere vMotion to migrate VMs between the hosts in the cluster By performing migrations, DRS ensures that VMs receive the required resources.
VM DRS Score
The VM DRS score is a new metric that measures the execution efficiency of a virtual machine. VM DRS score considerations:
• VM resource requirements fulfillment
• ESXi host resource capacity
DRS migrates VMs to the host that provides the highest score and ensures fair resource allocation to all VMs.
The VM DRS score roughly represents the percentage of time that a VM has its resource requirements satisfied over a sample period. The scores are filed into one of five buckets. These buckets are 20 percent ranges. A score of 80-100% indicates little to no resource contention.
DRS evaluates the current performance of a VM. Then, it runs what-if scenarios on compatible hosts to evaluate the cost of the VM migration and the potential performance. If the migration helps the VM run more efficiently, DRS migrates the VM to the new host.
Because of the new focus on the VM DRS Score, more migrations are likely after DRS is enabled. The cluster must be normalized to ensure that VMs have the required resources regardless of the individual host that they are placed on.
The Cluster DRS and VM DRS scores replace the balance deviation level in the cluster Summary tab.
Cluster DRS Score
The Cluster DRS score is a new metric that averages the individual scores of VMs. DRS focuses on maximizing the VM execution efficiency rather than on the host load balance in the cluster.
DRS Settings and Score List
On the vSphere 7 DRS portlet, you can access links for the DRS parameters and score list:
• Select VIEW DRS SETTINGS to view the DRS parameters and their values.
• Select VIEW ALL VMS to access the new VM DRS Score list.
VM DRS Score List
In vSphere 7, DRS provides a listing of the VM DRS scores and more detailed metrics for all cluster VMs.
A VM DRS score is computed from an individual VM’s CPU, memory, and network metrics. We can use these metrics to gauge the goodness or wellness of the VM. The metrics are useful in determining which hosts are experiencing resource contention and of what type.
Cluster DRS Score in Performance Charts
In vSphere, DRS includes a new metric in the performance charts. The new metric is the cluster DRS score.
This metric shows the cluster DRS score over the time span in the advanced performance chart.
An agent VM is a virtual machine that performs a specific function for the virtual infrastructure.
For example, an agent can act as a firewall or gather information about the virtual machines on the host.
Using the ESX Agent Manager (EAM) service, vCenter Server can deploy and manage agent VMs.
Agent VM Management in vSphere 7
In vSphere 7, agent VMs are pinned to an ESXi cluster. vSphere DRS can migrate agent VMs between ESXi hosts in the cluster.
If an ESXi host fails, vSphere HA always starts agent VMs first. This restart priority is nonconfigurable.
Resource Pool Shares
A resource pool is a logical abstraction of hierarchically managed CPU and memory resources. Shares specify the resource pool priority relative to other resource pools.
Resource pools can be organized hierarchically. The root resource pool is the topmost resource pool. The root resource pool includes the sum for all CPUs (in megahertz) and the sum of all the installed RAM (in megabytes) available in the compute environment (standalone host or cluster).
A child resource pool is used to allocate resources from the parent resource pool for the child’s consumers. A child resource pool cannot exceed the capacity of the parent resource pool. By creating a child pool, you reserve resources from the parent pool. In the example, the root resource pool is Cluster-A. Cluster-A has two child resource pools: the Engineering pool and the Finance pool.
Shares specify the relative priority or Importance of a resource pool. In the example, we want the finance VMs to have twice as much processor time as the engineering VMs. We can configure the Finance pool with twice as many CPU shares as the Engineering pool.
We can also specify shares per VM and distribute the resource allocation within the resource pool. VM shares are relative only to other powered-on VMs within the same resource pool. In the example, all VMs have an equal number of shares. However, the Finance VMs are entitled to about 67% of the cluster processor time. Each Finance VM gets about 33% of the entire cluster CPU resources, or twice as much as for engineering VMs.
About Scalable Shares
In vSphere 7, you can enable scalable shares to dynamically scale up and down the shares of sibling resource pools and maintain the resource allocation ratio.
When we enable scalable shares on a resource pool, all child resource pools maintain the original ratio of shares configured in the parent pool.
In the example, we enable scalable shares on Cluster-A to dynamically adjust the resource allocation of child resource pools. All VMs have an equal number of shares, so the shares configuration of the child resource pools adjusts to give every finance VM twice as much as the engineering VMs. The shares configuration of the child resource pools adjusts to 80% for the Finance pool, and 20% for the Engineering pool. Each finance VM gets 20% of the Cluster-A processor time, and the engineering VMs get 10%.
Enabling Scalable Shares
We can enable scalable shares at the cluster or resource pool level. The resource allocation of child resource pools becomes scalable.
vSphere DirectPath I/O
Using vSphere DirectPath I/O, VMs can directly access the physical PCI and PCIe devices connected to the ESXi host.
In earlier versions of vSphere, VMs identify the PCI devices by using the bus slot:
• The VM is bound to the host hardware.
• DRS does not migrate the VM using vSphere vMotion.
• vSphere HA does not restart the VM on a different host.
We configure PCI devices on the host to make them available for passthrough to a VM. With PCI passthrough, VMs can directly access high-performance graphics or sound cards, bypassing the VMkernel.
Dynamic DirectPath I/O
In vSphere 7, with dynamic vSphere DirectPath I/O, VMs can identify PCI devices by using their vendor and model names:
• A pool of PCI devices that are available in the cluster can be assigned to the VM.
•At power-on, DRS performs initial placement of the VM on any host that can provide the identified PCI device.
• We cannot migrate the VM using vSphere vMotion.
• vSphere HA can restart the VM on any host that can provide the identified PCI device.
Adding Dynamic vSphere DirectPath I/O Devices to a VM
We can add multiple dynamic vSphere DirectPath I/O devices to a VM:
• PCI pass-through devices
• Shared PCI pass-through devices, for example, NVIDIA GRID GPUs
Full memory reservation is applied to the VMs with pass-through devices.
To add a dynamic vSphere DirectPath I/O device to a VM, go to Edit Settings for the VM and follow these steps:
1. Click ADD NEW DEVICE and select PCI Device.
2. For New PCI device, select Dynamic DirectPath IO and click SELECT HARDWARE. A list of devices that can be attached to the VM appears.
3. Select one or more devices from the list.
I hope it has been useful to you. See you next!