To Raid or not to Raid

[holabr]

I have two 1TB drives in my new system. I am trying to decide if I should use Raid 1 for redundancy or use them as seperate drives and do nightly incremental backups. The Raid 1 option benefits, as I see it, are up to the minute redundancy and some improvement in read times. The seperate drives/incremental backup gives me point in time recovery capabilities but could result in a days lost work. Does anyone have any other input that would or could have an impact on the decision?

 

[sub.mesa]

Choice between RAID1 OR backups is simple: BACKUPS! Much more reliable.

Incremental backups done periodically (like every 6 hours) are possible to reduce the time files become stale (differ from the original).

If the data is truly important, consider a checksumming filesystem like ZFS.

 

[mv2devnull]

Yes, even if you would have RAID, you would have to backup.

With RAID1 you can survive a loss of HDD, but you want to have backups in case that the second disk dies too, before rebuild is complete. Besides, you can lose files due to other causes than loss of HDD, and there mirror won't help.

 

[RebateMonger]

Backups.

 

[FishAk]

Raid 1 has a performance boost over single disks, but it's not a backup solution. First backup, than RAID.

 

[sub.mesa]

@FishAk: as far as i know, only FreeBSD's geom_mirror with 'load-balance' or 'round-robin' algoritm would actually enhance RAID1 read performance. Virtually all other RAID1 implementations just show single disk read performance. Write performance would likely always be that of single disk, though also here RAID1 allows potential which no implementation uses.

RAID is great, but it's just a spec on how the data is spread over several disks. The actual implementations dictate what kind of performance potential is actually used and how reliable it is.

For example, RAID can increase reliability, but also decrease it. You are adding another layer to your chain of storage which can fail. And when it does, it will either temporarily or permanently prevent you access to your data. That can also happen without either disk failing, which makes a single RAID1 without backup much less reliable than many expect it to be.

Always go for a backup, is my advice. Though that still leaves the possibility to use RAID on either of them. But the less things the backup and original share, the better.

 

[pitz]

Yeah, the RAID-1 really helps performance (or at least it does here on Linux). But, one should still have an 'outboard' backup solution if they value their data (whether it be a backup to another server, or an external USB drive that's plugged in, etc.).

Why? Because it is possible that a virus or simply an errant user comes along and wipes out both the main drive, and the backup. Or a power surge wipes both out simultaneously. Or a water main break.

 

[Rubycon]

Raid and backups are different things altogether. RAID is for availability and backups are to protect your data from theft, disaster, hardware failure, PEBKAC, even sabotage! Backups should be offsite and onsite and preferably in at least two physical locations separated by 100 miles or more.

 

[7beauties]

In my humble opinion RAID is overrated. I've tried it on at least two occasions when I was upgrading my system. I thought that RAID 0 was supposed to double the read and write times and therefore I'd notice faster loading of programs. There's absolutely no difference; if there is, it's too minute for you to notice. The last time I tried it my hard disks drives were so hot that I could smell them! My room was so warm from the heat that I could close the heat register in winter. I bought passive heatsinks for my drives which have cooling fins that face out of the front of my 5.25" bays and they were very warm to the touch. I still can't believe that my hard disk drives didn't utterly fail from operating so hot. Eventually I disabled the RAID array, and repartitioned the drives individually and the heat was gone. Perhaps motherboard manufacturers have refined RAID, but I'll never go back to it. There are two other very important considerations you must take into account for RAID 0 or 1, and that is the expense of two drives for the storage space of one; two 1TB hard disk drives in RAID doesn't give you 2TB of storage. Bummer. And in the case of RAID 1 for redundancy, if any drive becomes corrupted by a malicious program or virus, then the other one faithfully copies the corruption as well. I would simply backup my files and third party programs onto a second drive. I hope that other people here who know more than I do weigh in on these important cons of going RAID.

 

[sub.mesa]

7beauties: please learn not to make conclusions before having the all the relevant facts.

HDDs power consumption is pretty much static; meaning that whether you let them work hard or not, the power consumption will stay about the same level. Most of the power consumption on HDDs is to maintain the spindle speed; in essence it is fighting friction of the rotating platters with the air inside (yes they are not vacuum like some think).

In other words, RAID cannot make your HDDs overheat. You likely use 7200rpm disks, which consume alot of power. That's why i recommend 5400rpm disks which use only half the power still get about 90-100% of the sequential performance. Normally 75% but since 5400rpm drives often have higher density platters this makes up for the rpm difference. You don't really need IOps performance from HDDs anymore if you utilize SSDs; so 5400rpm might be all you need in the future. This should make your heat issues go away.

Second, know that you are judging your motherboard's RAID drivers and your whole setup/situation; not RAID in general. If your speeds are low that's not the fault of the RAID specification, but rather your drivers, computer or specific product implementing RAID.

But i suspect you have been using RAID1 instead of mirroring: two 1TB disks in RAID0 *DOES* give you 2TB. So you might have actually used RAID1 (mirroring) instead of RAID0.

You're generally right though that RAID on windows sucks. But when you discover ZFS, you will see how RAID can be implemented in a smart and clever way. "RAID" is just with us because no advanced filesystems on windows exist that directly access multiple harddrives. Windows only got single-volume filesystems. In the future this will be considered obsolete and much more advanced filesystems will store your data. Did i mention ZFS? ;-)