Virtual File Server with Volume Larger than 2 TB

[jshapiro]

I have an application where I need to virtualize a file server where the data set could grow to over 2 TB. I know VMDK tops out at 2 TB. I am trying to weigh my options when allocating storage in my iSCSI SAN. Am I better off mapping a RAW LUN to Windows to GPT partition, or am I better off converting several VMFS formatted VMDK volumes into dynamic disks so I can extend the volume set in Windows? I am concerned for not only file server performance but also backup performance and of course capacity to hold the backups. I anticipate that I will have to do network-based backups if I work with RAW storage. That would also impact performance.

Would be curious to hear what others are doing.

[Gostev]

Dynamic disks would do the job, but then you lose instant file-level recovery capability (which is a real issue for file server). Is it a flat files list, or some folder structure? Because you could have multiple volumes (multiple VMDKs), but create volume mount points to all volumes on a single "main" volume. Thus, your end users will see all the folders in a single place, despite them pointing to the different volumes. I guess this might be the best solution, unless someone else has the better idea.

[MattG]

Couple of things to keep in mind:
- vStorageAPI only supports up to 1TB VMDK when using Hot/Add mode. If you use a VMDK > 1TB then you need to use Network or a physical Veeam backup server.
- Larger partitions can be a pain point when scanning with AV

When possible I prefer to keep my File Server LUNs < 1TB and then use MS DFS to point to them with a single namespace.

-MattG

[Gostev]

That's right, DFS is preferred (this is what we are using internally).