For home usage only, RAID5 good enough? or RAID6?

[Eeqmcsq]

Most of my research between RAID5 and RAID6 tells me that due to the size of HDDs nowadays and the possibility of an "unrecoverable read error" during a rebuild, RAID6 is the better choice because it can withstand 2 HDD failures and the data will still be available.

However, none of the articles say anything about HOME usage, whether or not RAID5 is "good enough" for an environment with light read and write usage. For example, if you had 5 x 2TB, would you go RAID5 for 8TB of storage space, but less fault tolerance? Or RAID6 for 6TB of space and more fault tolerance?

 

[SecurityTheatre]

Heh. Most people, including (apparently) Linus Torvalds' primary kernel development machine, run no RAID at all.

Given an anticipated failure rate of under 3% per year (for the first 4 years), it seems very unlikely to have a multi-drive failure....

 

[Cerb]

Heh. Most people, including (apparently) Linus Torvalds' primary kernel development machine, run no RAID at all.

Given an anticipated failure rate of under 3% per year (for the first 4 years), it seems very unlikely to have a multi-drive failure....

The problem is that drive failure can and does occur in multiples. In that, you repalce the failed drive, and a failure occurs during the rebuild where the other drives are experiencing far higher loads, during those hours of rebuilding, than they have have on them, before.

RAID primarily protects against you have to pay employees to do very little work, and/or not be able to serve clients, due to a disk failure. If you aren't running a business, then you have no need for RAID. Not saying you shouldn't use it, but there isn't a best RAID type for home users, nor worst RAID type for home users. As a non-business user, RAID is entirely a matter of (a) geek factor and (b) convenience. It's something you want, not something you need.

 

[Mushkins]

Obligatory "Raid is not a backup strategy."

RAID is for high availability, period. RAID 5 or 6 makes no difference, because who cares about drive failures when you're properly backing up your data. Linus Torvalds doesn't RAID his development machine because RAID is not a backup, you can guarantee he's doing something to back that code up properly.

High availability is totally moot for home use. The downtime you're trying to prevent with a RAID is meaningless in that environment, you're not losing employees time, productivity, and money if your movie collection is unavailable for a few hours while you swap and restore a dead drive.

 

[glugglug]

I'm using Stablebit Drivepool for my new system, with 4 drives.

Space efficiency wise it's the same as RAID10. Performance wise I think it is MUCH better for most use cases. And more importantly, I think it is a lot better data-safety wise because the individual drives each have a normal NTFS filesystem on them that can be mounted on another machine (with a subset of the files). Each file exists in its entirety on a single drive (well 2 drives with it set to make 2 copies of everything). You can set the replication level separately for each subdirectory, i.e. raise it to 3 for the most critical files.

 

[rsutoratosu]

I had multi drive failures, 6 wd green from 4 years ago, luckily 4 drive failed a week apart, else I be dead. I do raid 5 + backup. I just need a giant drive of storage.. not multiple little partitions..

 

[smitbret]

If you have a good backup system then RAID 5 is fine for home use, especially if you use one of the newer software RAID parity solutions like FlexRAID, unRAID or SnapRAID. With those, the data isn't striped so drive failure isn't completely catastrophic; if you lose more drives than you have parity discs, you only lose the data on the drive(s) that die. RAID 5 with backup means that if I lose 2 drives in my 5 drive setup, I just restore 2 drives worth of data from backup or rerip a few BR discs and recreate the array.

 

[KentState]

At home, I use RAID 5 with on-site and off-site backups.

 

[Emulex]

no raid at all is far more successful with consumer gear. IIRC windows 8 can throw a file based raid on top of jbod (no raid).

traditional raid-5/6 are not for consumer use. It will lead to many fails.

I don't even trust raid-5/50/6/60 with servers that are up 100% of the time. the rebuild time and its impact (without ssd caching) is too much.

Plus random i/o -> many raid-1 can dispatch more concurrency than one big raid-5/50/6/60

 

[aigomorla]

no raid at all is far more successful with consumer gear. IIRC windows 8 can throw a file based raid on top of jbod (no raid).

traditional raid-5/6 are not for consumer use. It will lead to many fails.

I don't even trust raid-5/50/6/60 with servers that are up 100% of the time. the rebuild time and its impact (without ssd caching) is too much.

Plus random i/o -> many raid-1 can dispatch more concurrency than one big raid-5/50/6/60

:T

i guess im the oposite of you.

If i had to do something with major fault tollerance.. i would RAID-1 and then RAID 0 the Raid 1 arrays for RAID10.

That way u would need to lose a LOT of HDD's b4 you started to have to worry about your data...
I know the rebuild time is a pain.. however id expect a reinstall time to be longer.

And i would do this over a dedicated controller with cache...


This will probably be very expensive... however the drives to setup a system like that would also be very expensive, and it will be a lot more fault tollerant then a standard raid1 array.

 

[Nec_V20]

I'd put it this way, you are more likely to destroy your hardware or set fire to your house because you decided that the cheap PSU you bought was good enough because you had blown your budget on the high class graphics card etc.

So if you do have a potential IED or as I like to call it "Al-Q Taliban Special" in your system you might want to replace it with a high quality PSU and then you will not have the likelihood of your hard drive being fried in the first place.

The one thing to consider about any kind of RAID is that all the drives will be spinning all the time. Do you really need that? If you are basically not doing anything which requires a huge contiguous space then why waste any storage space on RAID in the first place.

If you distribute your data then you can work on one hard drive and the others can go to sleep. When you work on some other data then that drive will spin up allowing the drive you were working on to spin down.

If you want your hard drive space to be quasi-contiguous then you might think about mounting your drives in sub directories. That is you have one drive which contains the directories and you mount the other drives to that.

You can then concentrate on making backups to other drives so that you won't lose your data if one drive fails.

Sure you have to think about it a bit more, however you won't be stressing all your drives out all the time.

 

[Cerb]

:T

i guess im the oposite of you.

If i had to do something with major fault tollerance.. i would RAID-1 and then RAID 0 the Raid 1 arrays for RAID10.

That way u would need to lose a LOT of HDD's b4 you started to have to worry about your data...
I know the rebuild time is a pain.. however id expect a reinstall time to be longer.

And i would do this over a dedicated controller with cache...


This will probably be very expensive... however the drives to setup a system like that would also be very expensive, and it will be a lot more fault tollerant then a standard raid1 array.

RAID 5 and 6 variants are parity RAIDs. RAID 1 and 10 aren't. RAID 10, however, will generally have a fairly quick rebuild time, and offers better random and sequential IO performance, at the cost of used space with many drives.

RAID 10 may or may not be more fault tolerant, though. That depends on which drives want to fail on you. While you might be able to handle up to n/2 failures, array loss could occur with as little as 2 drives, no matter how many are in the RAID 10, if they're in the same RAID 1. The thing is, it can be restored with pretty much a big sequential transfer from drive A to drive B. Parity RAIDs need to read from the other drives, make the parity or missing data, then write again, to just that location on the new drive, so even if the rebuild is fairly optimized, there's a lot of starting and stopped along the way, which tends to make it take forever.

 

[_Rick_]

I use a RAID 5, keep most of the stuff on there duplicated on another RAID 5, and the really important stuff duplicated to another RAID 1.

I don't believe in off-site storage much, because all my irreplaceable data has little real value, compared to what it takes to actually take out those three RAIDs.

Basically, losing the data would be similar to a nuclear accident, because a gigantic earth quake and tsunami have just devastated thousands of square kilometers. Sure it's noteworthy, but in the big picture I won't worry about that too much.

Of course, stuff like book keeping info for your business might best be served with an encrypted off-site backup, basically, stuff that your livelihood depends on.
Most of the other stuff is just nostalgia anyway.