RAID logitics question

[imojojojo]

Country: USA

Games, programs or workloads that it will be used for:  Live recording/streaming setup

Hello friends!

I'm putting together a rig that will be used for live recording in an ATA case. One of, if not the biggest concern is redundancy. If something goes wrong we need to be able to get back up and running as soon as humanly possible, less than a minute ideally. My plan at the moment is to have two software SSD's in a RAID 1 config so if drive 1 fails we can boot over to drive 2. I dont have any experience with RAID, however, so I dont know if logistically this would work.

I plan on using this ASRock motherboard and two of the classic Samsung SSD's. My question is simply if what I'm planning would work? I know you setup the RAID array in the BIOS, but when I come to a point when drive 1 dies what do I do to switch over to drive 2? Do I delete the array so its seen as 2 separate drives again? Is there a better way for fast and easy redundancy?

For clarification: My current plan is that if drive 1 fails, I simply boot into drive 2, since a RAID 1 array just copies the data. Am I wrong in thinking that it will be relatively easy to just boot into the 2nd drive? 

Any help is appreciated! I hope you have a nice day! :D

 

[Electronics Wizardy]

What os are you using?  Windows 10 doesn't support booting in software raid.

 

MOtherboard raid sucks, so id get a raid card, and boot fro mthat. That way a drive can fail with no reboot, and the os sees it as one drives.

 

But with how ssds don't fail ofte, Id probably not bother with raid here, and just get a single ssd(might as well go nvme here too)

[imojojojo]

2 minutes ago, Electronics Wizardy said:

What os are you using?  Windows 10 doesn't support booting in software raid.

 

MOtherboard raid sucks, so id get a raid card, and boot fro mthat. That way a drive can fail with no reboot, and the os sees it as one drives.

 

But with how ssds don't fail ofte, Id probably not bother with raid here, and just get a single ssd(might as well go nvme here too)

We'll be using Windows 10. A RAID card completely slipped my mind, thanks for the recommend!

I dont plan on it dying any time soon, but out budget has enough room for stuff like a RAID controller plus 2 SSD's and we'd rather be safe than sorry on this one. The ATA case will be moved around frequently and small stuff like a RAID card is a drop in the bucket, so having the ease of mind is worth the extra money to us.

Thanks for your help, I really do appreciate it!

[manikyath]

a raid array 'constantly' uses every drive in the array.

for completeness, the entire thing. typed out, because linking to wikipedia is rude:

(in this explalantion, i'm assuming each disk in the array is identical)

RAID0 (also known as striping): the data gets split into slices that are spread equally among all disks, greatly increases read/write speeds, and capacity is the total of all disks, but offers no redundancy for drive failure. one drive dead = all data gone.

RAID1 (also known as mirroring): writes all data to every disk so they all contain an identical copy, reads are distributed among the drives to increase read performance. total capacity of the array is the size of one disk. if one drive dies, the system wont even notice, you can just replace the broken drive (means depend on the specific controller) and the array will repair itself.

RAID5 (parity based): at least 3 drives in an array, you lose one drive worth of storage compared to the total capacity of all disks. that one drive worth of data is used for 'parity' information. essentially a mathematical calculation is used to create a set of data from which any one disk worth of data can be rebuilt. 

RAID6 (parity based): same as above but 4 disks minimum, and you lose two drtives worth of storage for two sets of parity, this allows for two failures to happen at once before the entire array is lost.

 

the theory lesson aside..

- booting from software raid is not a thing on windows. some linux sided things support it, but it's complicated to say the least.

- as @Electronics Wizardy so nicely implied: motherboard raid is not ideal. the upside of it is that it uses the system's resources for doing the raid operations so you dont need additional hardware, but it's not unlikely that bios updates will make you lose the array, doing a bios reset loses the array, the bios somehow resetting itself (you know.. dead cmos battery, something getting corrupted, modern mobo's switching to their backup bios, etc.) loses the array.

- a hardware raid card can be had fairly cheap, do make sure it's one that actually does the raid on the card itself, and isnt just something driver level.

 

if you can deal with the downsides of motherboard raid (y'know, have a regular backup, document bios settings so you can recover the array if necessary, etc.) and are on a tight budget, it's probably not worth the cost of a propper hardware raid card. if guaranteed uptime is a thing wherever this device ends up.. it's probably best to just toss some 4-port highpoint rocketraid card in there. dont forget to give it some airflow tho.

[Radium_Angel]

If this is critical work, why not go with something designed for critical work, like a server or high-end workstation, where all that stuff is already there and built to take abuse?

[manikyath]

2 minutes ago, Radium_Angel said:

If this is critical work, why not go with something designed for critical work, like a server or high-end workstation, where all that stuff is already there and built to take abuse?

from someone with experience in medical and other high-risk IT environments...

 

because they dont want to pay for it.

 

let's say i've personally configured a pair of fairly standard Win10 HP ProDesks with virtual machines, because a customer 'didnt have the budget' for a propper server with a propper server OS for their critical new application.

[Radium_Angel]

Just now, manikyath said:

from someone with experience in medical and other high-risk IT environments...

 

because they dont want to pay for it.

 

let's say i've personally configured a pair of fairly standard Win10 HP ProDesks with virtual machines, because a customer 'didnt have the budget' for a propper server with a propper server OS for their critical new application.

I'm right there with you, but really, if this is mission-critical work, do it right, or don't do it at all.

Of course, I also want a magical flying pony too, let's see which one happens sooner...