Use case scenario of Thick eager zeroed and Thick Lazy zeroed

[Ranjna Aggarwal]

Anyone who can tell me the Use case scenario of Thick eager zeroed and Thick Lazy zeroed?

[vmroyale]

VMware FT requires virtual machines to have thick-eager zeroed disks.

[Ranjna Aggarwal]

and any other example for this and and example for lazy zeroed

[rubensluque]

Thick eager zeroed is also used when data security is a concern. When you delete a VMDK the data on the datastore are not totally erased what acctually happen is that area is marked as available so OS can overwrite on that. When you create a virtual disk using the "Thick eager zeroed" on a datastore that had previous data the area will be totally erased (shred) and no previous data can be recovered by a person who has bad intentions.

[MKguy]

I can't find a source for that at the moment (think I saw it in a whitepaper), but I'm pretty sure that read requests against thin-provisioned or lazyzeroed blocks that haven't been written to on a VMDK will always return zeros and not the actual data of the physical disk this block is located on, so security should not be a concern.

As was mentioned already, eagerzeroedthick disks are required for FT and also for MS clustering (cluster-in-a-box and for a node's system disk). Besides that, eagerzeroedthick disks provide the best write performance from the very first write on.

But performance will be identical after a block has been written for the first time on a lazyzeroed disks.

[rickardnobel]

MKguy wrote:

I can't find a source for that at the moment (think I saw it in a whitepaper), but I'm pretty sure that read requests against thin-provisioned or lazyzeroed blocks that haven't been written to on a VMDK will always return zeros and not the actual data of the physical disk this block is located on, so security should not be a concern.

That is correct. There will never be an possibility to read old data, everything has to be erased ("zeroed") and it is only a question of when: eager = do it at creation = lots of write IOs to the SAN, could take a long time to create the virtual disk, but best performance. Lazy = do it when it is needed = somewhat slower write access at new locations.

Long ago there was a "thick" which was without any zeroing at all, but that is no longer possible to create.

[Ranjna Aggarwal]

Can you give me any use case scenario of Thick Lazy Zeroed?

[rickardnobel]

Ranjna Aggarwal wrote:

Can you give me any use case scenario of Thick Lazy Zeroed?

Thick Lazy Zeroed is fast to create and by that easy to the vSphere Admin to work with, e.g. a 1 TB disk is created in seconds. It does not have the risk to overwhealm the SAN with an extreme amount of write IOs (could be less problem with VAAI write-same). Most VM uses does not notice the somewhat lower performance at first write to a specific block. The advantage over Thin is also that it does allocate all space in the VMFS metadata and by that lower the risk of the datastore becoming full.

For these reasons it is also the default disk type.

[Ranjna Aggarwal]

But what is the use case can you give me any example for each type of disk provisioning method? Like Application A requires this or that.

[rickardnobel]

Since it is the default the use case is for almost all types of VMs. If your virtual machine writes a lot of data and has high performance demands, like a SQL server transaction logs, then a Eager Zeroed is better. If you want to save SAN space then Thin could be better, but with the risk of datastore overcommit.

[Ranjna Aggarwal]

Is this ok if i will compare the Thick Eager zeroed with Full Format and Thick Lazy zeroed with quick format?

[rickardnobel]

Ranjna Aggarwal wrote:

Is this ok if i will compare the Thick Eager zeroed with Full Format and Thick Lazy zeroed with quick format?

Not really actually. The Thick Eager is something like a full format on Windows 2008, but not a full format on Windows 2003. The behavior is different: http://rickardnobel.se/vmware-thin-disks-windows-formatting-options

[Ranjna Aggarwal]

But for windows 2008 that means we can say like this Eager Zeroed is like full format and Lazy Zeroed is like quick format.Am i right?

And thanks for this valuable Info. But why it is different in windows 2003 and windows 2008?

[MKguy]

No you're not, OS filesystem formatting of a storage device and eagerzeroed/lazyzeroed provisioning done by virtualization hosts are simply 2 different things. You can't compare them like that.

So what other piece of information are you still looking for and why is it so important? Pretty much everything has been said already.

[rickardnobel]

Ranjna Aggarwal wrote:

But for windows 2008 that means we can say like this Eager Zeroed is like full format and Lazy Zeroed is like quick format.Am i right?

I would say still not really. The point of the zeroing process from the VMware point of view is to guarantee that no old data stored on the SAN could be accessed through a new VM. This means the zeroing has to done, just sooner or later (eager or lazy).

Inside Windows it might be the same, but I have never seen it published if Windows does on-demand-zeroing for new files. With a full format in Windows 2008 and later it is however guaranteed that no old data remains.

IF Windows does indeed do a zeroing before new files are located and it keeps track on what has been and not zeroed, and if a full format would change this behavior, then it could be comparable, but unfortunately it is unknown to me.

And thanks for this valuable Info. But why it is different in windows 2003 and windows 2008?

Sometimes behavior changes in the operating systems. Many people using Windows 2003 and earlier might have thought that a full format actually deletes data, but it did not. Perhaps was this the reason to begin to really zero the disk in Windows 2008.

[mikejroberts]

Thick Lazy would be better if using Thin Provisioned LUNs on your storage array.  Eager zeroed will cause the blocks to be allocated on the array whether you end up using them or not.