Have an account?

  •   Personalized content
  •   Your products and support

Need an account?

Create an account

HyperFlex Boost Mode White Paper

White Paper

Available Languages

Download Options

  • PDF
    (1.2 MB)
    View with Adobe Reader on a variety of devices
Updated:April 17, 2020

Available Languages

Download Options

  • PDF
    (1.2 MB)
    View with Adobe Reader on a variety of devices
Updated:April 17, 2020
 

 

Introduction

HyperFlex Boost Mode increases the available storage IOPs of a Cisco HyperFlex (HX) cluster. It does this by increasing the number of vCPUs allocated to the storage controller Virtual Machine (VM) by four. The availability of this feature starts with Cisco HyperFlex Release 4.0.1b.

HyperFlex Boost Mode is a manual process that is supported for the limited set of configurations enumerated below. Customers are responsible for ensuring their clusters meet these requirements.

There are no license requirements for enabling the HyperFlex Boost Mode.

VMware CPU settings explained

In the settings of the VM, you can set CPU variables:

      CPU: Number of virtual CPUs (vCPUs).

      Cores: Number of cores per virtual CPU.

      Reservation: Guaranteed minimum amount of CPU cycles for the VM. If the VM is not using all cycles, other VMs can use the unused resources at that time.

      Limit: Maximum amount of CPU cycles for the VM.

      Shares: Depending on the amount of shares a VM has, when there is resource contention the CPU cycles are shared among the shares. Shares are a relative metric for allocating CPU cycles.

Configuration requirements

A HyperFlex All-Flash with nonboost mode configuration needs eight vCPUs with one core per socket and a reservation of 10,800 MHz.

The HyperFlex All-NVMe needs 12 vCPUs in nonboost mode.

To enable HyperFlex BoostMode, add four additional vCPUs to the controller VM.

For changing the number of vCPUs of the HyperFlex Controller VM, you will need the Virtual Machine, Configuration, Change CPU Count privilege in your vCenter account.

Requirements

      Supported hardware:

    HyperFlex All-NVMe

    HyperFlex All-FlashHXAF240 andHXAF220

      Hypervisor: ESX only

      HyperFlex Boost Mode requires the following number of vCPUs on the controller VM

    HyperFlex All-NVMe: 16

    HyperFlex All-FlashHXAF240 andHXAF220: 12

      CPU: The number of physical cores must be at least the new number of controller vCPUs. In the vSphere Client, click on the host, then Configure, then Processors, and verify “Processor cores per socket”. (See Figure 1.)

      Table 1 has a list of supported CPUs.

Verifying number of processor cores per socket in a HyperFlex ESXi host

Figure 1.            

Verifying number of processor cores per socket in a HyperFlex ESXi host

Enable HyperFlex Boost Mode

The implementation is like a rolling upgrade of the HyperFlex Controller VM (HX CVM).

Figure 2 shows the workflow to enable HyperFlex Boost Mode.

Workflow to enable HyperFlex Boost mode

Figure 2.            

Workflow to enable HyperFlex Boost mode

Configuration of HyperFlex Boost Mode:

Step 1.     Verify that the HyperFlex cluster is Healthy:

Go to HyperFlex Connect – Dashboard; the Resiliency Health should be “Healthy.”

Verifying the HyperFlex cluster is in a healthy state

Figure 3.            

Verifying the HyperFlex cluster is in a healthy state

Step 2.     Stop the HyperFlex CVM gracefully via: Shutdown Guest OS.

Graceful shutdown of HyperFlex Controller VM (CVM)

Figure 4.            

Graceful shutdown of HyperFlex Controller VM (CVM)

Note:     If the original HyperFlex system is HXDP Release 4.0.1a or lower, and you upgraded this HyperFlex cluster to later releases, then the HyperFlex Controller VM (HX CVM) is controlled via the vSphere ESX Agent Manager. This means that the agent always wants to have the HX CVM up and running. If you shut down the HX CVM, there is a possibility that the HX CVM will power on automatically.

Confirm shutdown of HyperFlex Controller VM

Figure 5.            

Confirm shutdown of HyperFlex Controller VM

Note:     A vCenter API call to shut down the Guest OS of the HX CVM is not possible.

Note:     When you put a HyperFlex node into Maintenance mode, you are unable to change the HX CVM CPU settings.

Step 3.     Wait until the HyperFlex Controller VM is powered off:

Verifying the HyperFlex Controller VM is shutdown

Figure 6.            

Verifying the HyperFlex Controller VM is shutdown

Step 4.     In HyperFlex Connect, you will see the current status changed to “Warning” for the HyperFlex cluster. Now the HyperFlex cluster is still Online, but not Healthy.

Verifying health of the HyperFlex cluster

Figure 7.            

Verifying health of the HyperFlex cluster

Step 5.     Edit settings of HyperFlex Controller VM:

Edit Settings of HyperFlex Controller VM

Figure 8.            

Edit Settings of HyperFlex Controller VM

Step 6.     After the controller VM is powered down, increase the number of vCPUs by four in the vSphere Client.

Click Edit Settings for the VM and the CPU value.

The new CPU value for a HyperFlex All-FlashCluster will be: 12 CPU

The new CPU value for a HyperFlex All-NVMe will be: 16 CPU

Changing the number of CPUs on the HyperFlex Controller VM

Figure 9.            

Changing the number of CPUs on the HyperFlex Controller VM

Note:     It is tempting to enable the CPU Hot Add for future usage. Don’t enable this feature, because, by doing so, you will change the vNUMA, which affects the performance of the HyperFlex cluster in a negative way.

Step 7.     Power on the HyperFlex Controller VM:

Powering on the HyperFlex Controller VM

Figure 10.         

Powering on the HyperFlex Controller VM

Step 8.     Go to HX Connect – Dashboard, and wait for the cluster to become healthy again.

Step 9.     Repeat steps 1-8 for each HyperFlex Controller VM in the cluster.

Note:     When the HyperFlex Cluster is expanded after HyperFlex Boost Mode is enabled, all new nodes require to have the above steps applied.

Disable HyperFlex Boost Mode

HyperFlex Boost Mode can be disabled by reducing the number controller VM vCPUs back to 12 (for All-NVMe) or 8 (for All-Flash HXAF240 and HXAF220) by following the implementation steps above. In Step 6, decrease the number of vCPUs by four.

Supported CPUs

In Table 1, you will see a list of CPUs. This list shows if the CPU supports the HyperFlex Boost Mode for a HyperFlex All-Flash or HyperFlex All-NVMe cluster.

Table 1.        Supported CPUs

HX CL PID

HX Boost All-Flash

HX Boost All-NVMe

Cores

Speed

HX-CPU-I4208

No

No

8

2.1

HX-CPU-I4210

No

No

10

2.2

HX-CPU-I4214

Yes

No

12

2.2

HX-CPU-I4216

Yes

Yes

16

2.1

HX-CPU-I5215

No

No

10

2.5

HX-CPU-I5217

No

No

8

3

HX-CPU-I5218

Yes

Yes

16

2.3

HX-CPU-I5220

Yes

Yes

18

2.2

HX-CPU-I6226

Yes

No

12

2.8

HX-CPU-I6230

Yes

Yes

20

2.1

HX-CPU-I6234

No

No

8

3.4

HX-CPU-I6238

Yes

Yes

22

2.1

HX-CPU-I6240

Yes

Yes

18

2.6

HX-CPU-I6242

Yes

Yes

16

2.8

HX-CPU-I6244

No

No

8

3.6

HX-CPU-I6248

Yes

Yes

20

2.5

HX-CPU-I6254

Yes

Yes

18

3.1

HX-CPU-I8260

Yes

Yes

24

2.4

HX-CPU-I8268

Yes

Yes

24

2.9

HX-CPU-I8270

Yes

Yes

26

2.6

HX-CPU-I8276

Yes

Yes

28

2.3

HX-CPU-I8280

Yes

Yes

28

2.7

HX-CPU-I6240M

Yes

Yes

18

2.6

HX-CPU-I8260M

Yes

Yes

24

2.3

HX-CPU-I8276M

Yes

Yes

28

2.2

HX-CPU-I8280M

Yes

Yes

28

2.7

HX-CPU-I5215L

No

No

10

2.6

HX-CPU-I6240L

Yes

Yes

18

2.6

HX-CPU-I6238L

Yes

Yes

22

2.4

HX-CPU-I8260L

Yes

Yes

24

2.3

HX-CPU-I8276L

Yes

Yes

28

2.2

HX-CPU-I8280L

Yes

Yes

28

2.7

HX-CPU-3106

No

No

8

1.7

HX-CPU-4108

No

No

8

1.8

HX-CPU-4110

No

No

8

2.1

HX-CPU-4114

No

No

10

2.2

HX-CPU-4116

Yes

No

12

2.1

HX-CPU-5115

No

No

10

2.4

HX-CPU-5117

Yes

No

14

2

HX-CPU-5118

Yes

No

12

2.3

HX-CPU-5120

Yes

No

14

2.2

HX-CPU-6126

Yes

No

12

2.6

HX-CPU-6130

Yes

Yes

16

2.1

HX-CPU-6132

Yes

No

14

2.6

HX-CPU-6134

No

No

8

3.2

HX-CPU-6136

Yes

No

12

3

HX-CPU-6138

Yes

Yes

20

2

HX-CPU-6140

Yes

Yes

18

2.3

HX-CPU-6142

Yes

Yes

16

2.6

HX-CPU-6144

No

No

8

3.5

HX-CPU-6146

Yes

No

12

3.2

HX-CPU-6148

Yes

Yes

20

2.4

HX-CPU-6150

Yes

Yes

18

2.7

HX-CPU-6152

Yes

Yes

22

2.1

HX-CPU-6154

Yes

Yes

18

3

HX-CPU-8153

Yes

Yes

16

2

HX-CPU-8158

Yes

No

12

3

HX-CPU-8160

Yes

Yes

24

2.1

HX-CPU-8164

Yes

Yes

26

2

HX-CPU-8168

Yes

Yes

24

2.7

HX-CPU-8170

Yes

Yes

26

2.1

HX-CPU-8176

Yes

Yes

28

2.1

HX-CPU-8180

Yes

Yes

28

2.5

HX-CPU-6134M

No

No

8

3.2

HX-CPU-6140M

Yes

Yes

18

2.3

HX-CPU-6142M

Yes

Yes

16

2.6

HX-CPU-8160M

Yes

Yes

24

2.1

HX-CPU-8170M

Yes

Yes

26

2.1

HX-CPU-8176M

Yes

Yes

28

2.1

HX-CPU-8180M

Yes

Yes

28

2.5

Verification of HyperFlex Boost Mode.

You cannot verify if HyperFlex Boost Mode is enabled via HyperFlex Connect, HyperFlex API, HXCLI, STCLI, or Cisco Intersight™.

Manual verification via vCenter is needed. You must verify that all of the HX CVMs of the HyperFlex cluster have the right number of vCPUs.

If you have a HyperFlex All-Flash cluster, in the vCenter settings of the HyperFlex CVM you must see:

Verifying that HyperFlex BoostMode is enabled on a HyperFlex All-Flash cluster

Figure 11.         

Verifying that HyperFlex BoostMode is enabled on a HyperFlex All-Flash cluster

If you have a HyperFlex All-NVMe cluster, in the vCenter settings of the HyperFlex CVM you must see:

Verifying that HyperFlex Boost Mode is enabled on a HyperFlex All-NVMe cluster

Figure 12.         

Verifying that HyperFlex Boost Mode is enabled on a HyperFlex All-NVMe cluster

Conclusion

To gain more performance on your HyperFlex All-Flash or All-NVMe cluster, you can enable HyperFlex Boost Mode by adding four vCPUs to each controller on the nodes.

HyperFlex Boost Mode is supported when the number of physical cores is at least the same as the new number of the HyperFlex storage controller VM vCPUs.

HyperFlex Boost Mode does not require any additional licenses.

Our experts recommend

Learn more