Dynamic or Static VDI for RAID?

HeartOfGermany · March 22, 2018

Hi, I want to use my Windows 10 PC as a RAID Server. It needs to work with a GPU, so PCI passthrough is no solution. No Unraid or Freenas so to speak.

Windows 10 RAID tools are to propietary and incomplete for my opinion. My solution is as follows:

Each HDD for the RAID gets mounted in Windows 10 and a virtual HDD (filename.vdi) gets placed on it. This later can be read inside a vm as a HDD and can be used for RAID and NAS. Is it better, to use a fixed file size for the HDDs or should I use a dynamic size for the HDDs?

I already thought about file size limit. So there is only NTFS as option, but no problem - it can even be mount read/write in almost all linux distros, so data recovery is always possible, if it boils down to that specific problem. Of course every extra layer is not optimal, But without virtualization of the Windows host there is no other way to use MDADM which is an absolute must have for me.

So, dynamic or static .vdi files? Dynamic has the advandage to see, how much data realy is on. If the .vdi is static, the hdd will always show as full in windows.

Apart from that I actually do not have any more questions. All seems very clear from that on.

ElSeniorTaco · March 22, 2018

I want to clarify, you want to take a few physical hard drives, and drop a vdi file on each one, and then use a vm to connect all the different vdi's together and make a raid drive out of the connected vdi files.

And your question is, if you do this, should you use dynamic or static vdi files.

If this is indeed the question. I think its possible that it wont matter which one you use,

Since the vm software handles the vdi files, I believe your VM wont notice anything special about the drive.

but as I have never heard of someone trying this, you might have to just try it and find out.

Also you might want to try the idea of recovering with any other linux device

HeartOfGermany · March 22, 2018

Hardware RAID Cards are very expensive and apart from the BBU and less ressource usage of the host, there is not much advantage. MDADM is a beutyful and very easy to use tool with any features I could dream off.

Hardware RAID would be the preferred method, if it was not that expensive. Also, to recover the data in case of card failure, I have to find a compativle/identical card - or better, buy 2 in the beginning. Which doubles the cost just for the ability to recover data if a card fails. I will use a cheap SATA Hub to connect all the drives to the pc. This is a very dirty method but should work perfectly fine. It is unlikely, that a controller failure nukes the drive. It only can write garbage before failing - which a hardware raid very well can do aswell.

My method is a bit more risky however saves tons of money. Also I had fun modding the PC-Case. Full I/O (card reader, 2 DVD Drives, 9x3,5'' hot plug bay [fully DIY!], up to 5x2,5'' for HDD and the 1TB system SSD, the GPU barely fits the case - litterally jammed in! :D)

I like to experiment a lot, which should be obvious by now.

I do not know the difference between static or dynamic vdi files. I could imagine, that it boils down to performance and wear leveling. I mean, if I write and delete files and only fill and kill the first 10% of the drives again and again I would propably destroy the first part of all drives, while the last parts are barely if ever used. Just what I think could be the thing.

Well shit, I'll just use the static drive - just imagine if windows writes on garble and the VM encounters criticall write errors because the space is no more free. So I will see, what exactly happens, when the file is 100% full from the beginning. Seems less risky.