The following best practices are recommended for the PowerFlex system:
- Ensure that the minimum number of nodes for production workload is seven, that means four storage only nodes and three compute only nodes for a PowerFlex two-layer configuration in a VMware environment. A minimum of three compute only nodes in an ESXi cluster is recommended to allow for HA/DRS.
- Homogenous node types are recommended for predictable performance. With homogeneous compute nodes and homogeneous storage nodes, compute and storage do not need to be the same.
- Change the passwords for all the default accounts.
- Verify that the PowerFlex rack/appliance is compliant to the latest PowerFlex RCM. For more details, see PowerFlex specification sheet.
- Enable the secure network protocol options only. For example, HTTPS and SSH.
- Ensure that the number of SDS threads is set to 12 for all the SDS nodes.
- Separate the management and data traffic from the production application traffic. You can provide this separation by using the VLANs.
- Separate the VMware vSphere vmotion traffic from the application traffic according to PowerFlex documentation.
- Disable the Intel c-state and p-state driver in the Linux grub setting to comply with the performance best practices and rebuild the grub boot image:
- It is recommended to use noop for Linux I/O scheduler on SDC hosts when SSD drives are used. To make the changes persistent through boot add elevator=noop to GRUB_CMD_LINUX line and rebuild the grub boot image.