The following summarizes Dell EMC best practices for VMFS datastores and virtualized ASM disks when deployed on ScaleIO:
- If using traditional VMFS datastores, create volumes that use the entire ScaleIO volume.
- Regardless of the VMFS volume/datastore configuration, a minimum of four ASM disks per diskgroup is still recommended to ensure that the guest OS has sufficient I/O queues to provide adequate performance.
- Dedicate one or more datastores for ASM use. Do not mix non ASM files into datastores used for ASM virtual disks.
- Add Scale IO volumes/VMFS volumes to extend datastores. Additional VMFS volumes will provide additional I/O queues to ESX.
- If multiple ScaleIO volumes are used for the ASM diskgroups, create a consistency group if array based snapshots, clones or replication will be used.
- Spread database disks (VMDK or pRDM) evenly across all available SCSI controllers in the VM.
- Use paravirtual (PVSCSI) adapters where available.
- If using RAC, the multi-writer flag must be set in the VMX file. (See VMware KB 1034165), and the VMDK must be selected as Eager Zero Thick.
- If using UDEV rules in Linux on VMware, ensure that the disk.EnableUUID=true directive is added to the VMX file of each VM that will access ASM.