So, I have a customer wanting to use a 750 series SSD as a boot and main program drive. No big deal. Then he wants to have a RAID5 with 3 6TB drives in it. I can't move up to RAID10 though as he can't afford another drive. Is it possible to run the PCI-E SSD and a RAID5 together? Would it be best to do a software RAID or maybe get a RAID controller? All input is appreciated.
It's possible to run that combination in software RAID. I run a 950PRO NVMe drive for boot and 4 850PRO drives in RAID5 for data myself.
You'll need to set the BIOS to run in RAID mode rather than AHCI. Windows doesn't like it if you switch and will refuse to boot, so you either need to do that before you install Windows or after you've backed everything up so you can re-install Windows.
I can't walk you through the BIOS, but a quick look in the motherboard's manual should give you all the info you need on that. The abbreviated version is as follows :
Swith the IDE controller from AHCI to RAID, save and reboot, during the boot go into the BIOS' RAID controller (usually with the ctrl-i combination), create a RAID 5 array and assign the 6TB drives as RAID disks.
Then reboot and install the OS on the SSD.
You can disconnect the 6TB drives when installing Windows on the SSD, and I wholeheartedly recommend doing so. Windows has a tendency to put the boot manager on the wrong drive, which would cause the OS to fail to boot if the RAID array ever dies or is removed for some reason.
After the install, just plug them back in. You can even do this when the array already contains data. As long as you unplug all 3 and do so while the PC is off, the RAID controller won't go into alarm or anything but it will remember how they were plugged in. If you plug them in again in the right position, the data will all still be there. I'd recommend puling the power cables rather than the data ones, that way you can avoid accidentally plugging the wrong drive into the wrong SATA port.
Personally I've never had reliability issues with such a RAID setup in my PCs. Had a couple of such RAID0 arrays in the past 5 years (when SSDs were still too slow to my liking), now this RAID5 since a month or so. I've also set some up for customers. Not a single array has ever failed on me.
The only time I actually resorted to using a proper RAID card is when I built my NAS, and even then I flashed it to HBA instead of RAID so that Freenas could do a software RAID.
As for the RAID5 vs RAID10 dilemma, in a 4-drive scenario I would actually pick RAID6 over RAID10.
With 4 drives in an array, there are 6 possible scenarios of a 2-drive failure (drives 1+2, 1+3, 1+4, 2+3, 2+4, 3+4).
With RAID6 the data is striped in such a way that any of those failures can be recovered from without data loss.
With RAID10 drives 1 and 3 will be mirrored, as are 2 and 4. So the 1+3 and the 2+4 scenario will result in complete data loss.
I know, I am promoting 4-drive RAID6 but using a 4-drive RAID5 myself, but there's a reason for that. If you need to rebuild an array after a drive failure/replacement, HDDs get worked really hard for an extended period of time. So a second drive failing during a rebuild is a likely occurence. My array consists of SSDs, so the rebuild takes a lot less time and due to the lack of moving parts it doesn't actually put stress on the drives.
On top of that the 850PRO is known to be extremely reliable, so I would be amazed if one ever were to actually fail on me.
One more tip :
One thing that I always do with new HDDs that are intended for a RAID array, is stress-test them for a couple of days before actually installing them in the PC. (actually I do it with any new HDD regardless of its intended purpose)
Each techie has their own tools and methods for that. I usually plug the drives into another PC as single drives, fire up CCleaner Portable (copying the folder of CCleaner Portable a couple of times allows you to run several instances of the program, so you can do all drives at once) and let CCleaner do a full 35-pass wipe of the drives.
That should give your 6TB drives a good workout for a couple of days. If a drive survives that, it'll usually last until well past the end of the warranty. If it doesn't survive it, the drive wasn't going to last long anyway.
Por Que.
True that. But if that's what the customer wants, that's what the customer gets.
At most, one can only suggest the better options, which IMO would be no RAID at all or a 4-drive RAID6
In a software raid, maybe. You totally should never do that though, way too many things could go wrong and you would probably be limited to the speed of the hdds.
I would really advocate you talk him into a raid 6 or raid 10, as raid 5 with drives that large is asking to lose data. It is very hard on drives to rebuild an array after a drive failure, and large drives compound the problem. There would be a large percentage change that rebuilding a raid 5 with drives that large would result in a second drive dying during recovery, and thus all data being lost. Raid 5 is a bad technology for drives that large, 6 or 10 is the only acceptable solution if you honestly care about your data with a 6tb drive. From there, you can go hardware raid controller and potentially see a tiny performance increase, but if attaining proper raid 6 or 10 means a controller doesn't fit in the budget, go for the raid 6 or 10 instead.
Or just use 1 drive and pay for an online back up service to actually probably better secure the longevity of his data
Well he should be doing an offsite backup regardless, but a single drive is far less capacity and can offer less performance than either a raid 5, 6, or 10. If he is looking at three of the six terabyte drives, I think a single six or ten terabyte drive probably won't satisfy his storage needs.
Is there really that big a speed increase for RAID 5? I mean it's still very large mechanical drives and all that
unless I'm reading it wrong RAID 5 is about a 25% write performance increase
Raid 5 can be better or worse than the figures presented depending on the drive and the caching and all the other processes that go on. I do agree, performance isn't as strong of point. There is definitely an advantage over a single disk, but its nothing to write home about. The reliability of raid 5 and 6 is something to really note.
He should always of course be doing a regular off-site backup, but there are also advantages to being able to rebuild your array locally. I mean a 6tb drive going offline is a lot of data to restore, and fucking hell I do not want to sit around downloading all that data when you could rebuild the array much faster.
This a great but long read over raid and which raid levels we should move towards: http://queue.acm.org/detail.cfm?id=1670144
Figure 6 under the heading "Comparing Raid Reliability" is a great illustration of the chance of failing to rebuild a raid 5 array vs rebuilding a raid 6 array when using what was predicted to bleeding edge drives, and actually using failure rates that are lower than what is being observed in industry.
I would recommend reading this, maybe sending it to your customer, and then making a hard sell case for raid 6. There really is a signifigantly higher chance of him losing all his data if he employs a raid 5, and frankly I would argue there is no excuse to not be using a raid 6, zfs raid z2, or even preferably a zfs raid z3 when using drives that are greater than 2TB in size, let alone 6TB cutting edge drives.
Holy tamales!!!!
What a great response from you guys!!! Thank you so much. I will definitely recommend this to the customer. Hopefully we can swing this. I have little experience in this and am looking forward to trying to get this under way.