New disk for DFS replication

[Gimbo58]

I have to create a new disk for VM machine. In this disk will be folder replicated trought DFS. Which kind of disk is better to create?

Thin , thick or thick eagerly zeroed?

[n00p · Accepted solution]

Thick lazy zeroed disks offer better performance and security while eager zeroed disks offer faster creation speeds. These disks reduce latency by preallocating physical storage on a disk. Thin disks optimize disk efficiency by allocating disk storage space between multiple users.

[Gimbo58]

Thank for your quickly response

I'll use "Thick lazy zeroed" option

[markey165]

Further to the answers above, I would argue the correct answer to this question is actually "......it depends".

The only requirement mentioned was a disk who's content would be replicated via DFS-R. There isn't enough information on things like "rate of change", performance requirements, or even what type of data is being stored on this disk to provide a correct answer.

If this disk is storing application data for example, or the data has high performance and low latency requirements, then yes, Thick Provisioning would be the optimal choice, particularly eager zeroed.   However, the downside with Thick Provisioning (of either type) is that it will consume 100% of the space allocated to the disk, even if the disk is only 10% utilized.

On the other hand, if this dataset doesn't change very often, or doesn't have high performance, low latency requirements, then a Thin Provisioned disk would be more efficient in terms of space consumed.  The caveat with Thin Provisioning is to be careful when over provisioning the storage. This is when the space allocated to disks on a VM (or VMs) is larger than the space available, but as long as you maintain sufficient overhead and monitoring, this shouldn't be a problem.  There is a performance overhead on writes of course, but if you only have few users writing the odd word document here and there, then the difference is negligible, and Thick Provisioning would probably be a waste of storage.

Just to add, if your datastores are on a Disk Array, some more intelligent arrays will reclaim unused space on a Thick Provisioned disk, typically by dedupe and compression, so you could thick provision everything, and on the back end, that lost space is actually recovered by the array.

Hopefully that provides a more complete answer 🙂

[Gimbo58]

Ok, very intersting

Some more information. This disk is dedicated to store files product by a CAD application (Inventor) and DFS sincronize this disk with another disk in a remote site

A part of DFS on this disk there are a lot of writing / reading activity from different users and (I don't know if it's important) every night Veeam execute a backup reading the full disk
I don't know if these informations are sufficient, basically I need good performance

[markey165 · Accepted solution]

There is a good article here which summarizes the pros and cons of Thick vs Thin disks

https://www.ubackup.com/enterprise-backup/thin-provision-vs-thick.html

It really depends on the individual requirements. Personally, I typically use Thick provisioned disks for high performance applications such as Databases (eg MS SQL, Oracle, MySQL), where performance is critical, but for an end user filestore I would opt for Thin Provisioned disks in the first instance as it is more space efficient. The performance difference will be small and in reality users probably won't notice the difference.

If you later wish to switch to thick, you can simply storage vMotion the VM and convert the disk to thick non disruptively, so you're not locked in if you wish to change.

Regarding Veeam, again the answer will vary depending on your setup, and whether you're using vcenter integrated backups or Agent based backups, but typically Thick provisioned disks will see a slight overhead over Thin disks.