Dynamic Striped VMDK and snapshot performance

[JoshuaPostSAMC]

We are preparing to setup some SQL Always On clusters in VMware and planning out disk configuration. The other Veeam administrator mentioned that they had seen better performance with Veeam and snapshot removal impact by doing multiple small VMDKs and creating them as Dynamic Disks in Windows and striping them. For example, we need 3 TB of storage for SQL, so creating 6x500GB VMDK. This originally came about before 5.5 was out and we could create a 2TB+ VMDK, but does it still make sense today? Should the impact to a VM be less when completing multiple disks for snapshot removal, or a higher risk or issues because one VMDK may take longer to remove the snapshot than the others and cause problems?

I also have concerns about the added complexity of dynamic disks vs basic, but that question may be better asked in a VMware forum.

[JoshuaPostSAMC]

Note: We are on ESXi 5.5 today. I understand that the snapshoting process is improved in 6.0 and may make this less of an issue, but would still like to hear thoughts.

[Delo123]

Not sure regarding snapshotting (I assume multiple vmdk's are actually a bit worse) but for restore purposes I would definitely split up the VMDK in chunks so you can do parallel restores (and in case of them failes it's better to retry 500GB than 2TB++).
Btw dynamic disk is not "really" striping, but a bit safer if you ask me (data retrieval in case of emergency) and also you can expand it later (both is rather impossible with real striping aka raid 0).