Should I use RAID?

[adzam]

I am trying to setup a fast, stable VM environment. Currently I have Windows 10 and programs on a Kingston HyperX 3k 120GB SSD and storage on a WD 1TB HDD. I plan on running Windows 10 as the host and at least one distro of Linux as a guest. A bunch of different storage scenarios have been going through my head and I am unsure which would be best for performance and stability.

Current system specs are:

• Gigabyte GA-Z97X-UDH5-BK (Marvell and Intel SATA chipsets)
  • Marvell supports RAID 0 and 1 (2 drives)
  • Intel supports RAID 0, 1, 5, and 10 (6 drives, 4 max on RAID 10)
• Intel i7-4790k 4.00GHz
• 32 GB RAM

These are some different scenarios I have been considering:

1. Add a second SSD, Keep host/programs on existing SSD and the other for VM's.
2. Add a second SSD and configure them in RAID 0. Run host and guests on array.
3. Add 3 more SSD's in a RAID 5 or 10 array. Run host and guests on array.
4. Add a Samsung 950 Pro 256GB M.2. Use that for host/programs, use existing SSD for VM's.
5. Do nothing. Keep host/programs on SSD, VM's on 1TB HDD.

The machine is primarily used for trading stocks/futures, web development and your normal usage (Office, web browsing, videos, etc).

I have already purchased 3 more Kingston SSD's. Even though I am itching to use them I can always return them if I don't need them or decide the M.2 is the best route. On a side note, I am not sure if my graphics card is using too many PCIe lanes to achieve full speeds on the M.2.

Thanks in advance for any input and suggestions!

EDIT: Option 4 is out. I am having trouble finding information, but it appears that my MB doesn't support M.2 Gen 3 so I would not be able to benefit from the full speed capabilities of the Samsung M.2.

[mpack]

None of the above, except maybe buy an SSD for the VMs to live on if you're short of space on disk 1.

For data safety, IMHO nothing beats a proper backup on external storage, and no kind of lazy high techiness substitutes for it.

And avoid snapshots and other differencing schemes like the plague. Of course with frequent backups the worst that can happen is you'd be inconvenienced, but then again I don't appreciate being inconvenienced!

[scottgus1]

A RAID array is used for continued availability if a drive (or two depending on the RAID number) goes south on you. We have RAID 1's in our server and in my workstation, and I have had drives go out, and I have been able to keep running until I could shut down and fix things. But RAIDs aren't backups. What is accidentally deleted off one drive in the array gets deleted off the other ones too. RAID 5's are known to stress still-running drives while rebuilding for a failed drive and sometimes kill off another drive during the rebuild, thus killing the whole array. RAID 0's have no redundancy for availability, and are for greater data bandwidth when drives were slower. A single SSD may out-perform a RAID 0 platter array.

As to performance, if you do decide to go with a RAID for availability in the event of hardware failure, just look up the RAID types and pick the one that works for your situation. The guest does not need to be RAIDed, just the host drive the guest's folder is on.

Virtualbox guests can be backed up two ways. If you are able to shut down the guest, use this method Moving a VM re-interpreted as "Backing Up a VM". If you want the guest live while backing up, get in-the-guest-OS live backup software and back up to a shared folder on the host network. I have read this on the internet: "Anyone who does not have at least two off-site backups of data does not value the data."

[adzam]

Thank you for your replies!

None of the above, except maybe buy an SSD for the VMs to live on if you're short of space on disk 1.

I am leaning toward just adding a second SSD and possibly dedicating it to my main guest with raw disk access.

For data safety, IMHO nothing beats a proper backup on external storage, and no kind of lazy high techiness substitutes for it.

I agree 100%!

A RAID array is used for continued availability if a drive (or two depending on the RAID number) goes south on you. We have RAID 1's in our server and in my workstation, and I have had drives go out, and I have been able to keep running until I could shut down and fix things.

I have used RAID 1 in all my servers, but never in a workstation. Lack of downtime due to drive failure is definitely a plus, especially if a drive happened to fail in the middle of a futures trade. The possibility of that happening is nothing I really considered until reading your post.

As to performance, if you do decide to go with a RAID for availability in the event of hardware failure, just look up the RAID types and pick the one that works for your situation. The guest does not need to be RAIDed, just the host drive the guest's folder is on.

I think RAID 10 would probably be best for me if I decide to go with RAID.

Virtualbox guests can be backed up two ways. If you are able to shut down the guest, use this method Moving a VM re-interpreted as "Backing Up a VM". If you want the guest live while backing up, get in-the-guest-OS live backup software and back up to a shared folder on the host network. I have read this on the internet: "Anyone who does not have at least two off-site backups of data does not value the data."

Thank you for the backup information and link! It will definitely come in handy.

[mpack]

Do not use raw disk access. Everything you know about the reasons why you might have wanted to do that - are not only wrong, but very very wrong.

Create a normal VDI on the SSD.

KISS!

[adzam]

Do not use raw disk access. Everything you know about the reasons why you might have wanted to do that - are not only wrong, but very very wrong.

Create a normal VDI on the SSD.

KISS!

Thanks for the advice! I will stick with a normal VDI.

[mpack]

I just want to correct myself in a minor way: when I say create a normal VDI on the SSD I do of course mean to create a normal VM folder on the SSD, of which of course the VDI will be a part.

Some people think they have to have the VM control files (amounting to a few K) on the main system drive, and a bare VDI in some random folder on the second drive. That's a bad arrangement as eiher the control files or the VDI tend to get forgotten during backups or moves, then nasty things happen (e.g. people who encrypted their drive lose the whole lot if they lose the .vbox). It is infinitely better to locate the entire VM in one convenient folder on the secondary drive.

When you create VMs, VirtualBox will create them by default in your "C:\VirtualBox VMs" folder. You can either leave it like that and move the VM after creation (FAQ link for that was given to you earlier), or you can change the default machine folder for newly created VMs in the manager preferences.