RAID 5+hot spare or RAID 6 for my new 8x2TB array?

X

XS Janus

Limp Gawd
Joined
May 7, 2008
Messages
416
I'm in the early stages of building my new NAS.
Currently I have a 8x1TB RAID 5 array, but it's almost full.:eek:

I will probably end up getting the new LSi 9260 card for this array, but I am curious to see what you guys think of these two raid sets.

What would be better to have in case one of the drives decides to die on me?
And for what specific reasons are each raid sets meant to be used?

Thanks for all the input, you'll have on this subject.:)
 
raid-5 will offer you faster rebuilds, as there's only 1 parity set to recalculate to, or from upon a drive failure. the drawback to this is that during the rebuild, any additional bit errors upon array rebuild (like when the hot spare is being incorporated, for example) will cause data loss on the entire array.

raid-6 is specifically to combat this raid-5 issue. however, because you're recomputing two parity sets, the rebuilds take longer. raid-6 is considerably more XOR intensive than raid-5. a half-assed RAID-6 controller will definitely perform very poorly. Even a decent RAID-6 controller will perform significantly slower than the same RAID controller doing RAID-5, instead.

your disk overhead will be the same, either way... you'll lose 1 spindle to the raid-5 parity drive, and 1 to the hot spare. on the raid-6 implementation, you lose two spindles worth of space to just the parity. if you have a hot spare on the raid-6 array, then you're losing 3 spindles worth.
 
I would go with RAID 6 over the RAID 5+1. With the RAID 6, you would still maintain your data if another drive were to fail while rebuilding the array after a single drive failure. With the RAID 5+1, if another drive fails during the rebuild (unless it's the one that was your hot spare) you're stuffed. Especially if you rely on RAID redundency and don't have proper backups.
 
sabregen covered everything I believe. At the amount of data with a 8x2tb array I think it's better to run a raid 6 for the safety of the extra parity set. IMHO, no matter what sort of backup you have with that amount of data, it's going to take some time and effort to do a restore and it's better to just run an extra parity drive (w/ hotspare...but that's just me)
 
Also, for decent hardware based card, read performance will be nearly identical for an 8 disk RAID 6 vs a 7 disk RAID 5 with hot spare (8 disks). Write performance will favor the RAID 5+1 by about 15-30 MBps based on the RAID engine.
 
I must say read/write performance is not a big issue for me this time.
When I build the thing I will definitely do some testing of the Raid levels mentioned to see how they behave. :)

the drawback to this is that during the rebuild, any additional bit errors upon array rebuild (like when the hot spare is being incorporated, for example) will cause data loss on the entire array
Click to expand...
I'm curious to learn what would cause this to happen? the drive not being "up to sync" with the array sort to say, or something else...
How big of an issue is this in reality? I know it could happen, but what are the factors that would increase the risk of this happening?

thanks!:)
 
XS Janus said:
I'm curious to learn what would cause this to happen? the drive not being "up to sync" with the array sort to say, or something else...
How big of an issue is this in reality? I know it could happen, but what are the factors that would increase the risk of this happening?

thanks!:)
Click to expand...

I believe sabregen is referring to Unrecoverable Read Errors (UREs), or at the very least URE's can cause what he's describing. A good read up on them: http://blogs.techrepublic.com.com/datacenter/?p=1752
which links to http://blogs.zdnet.com/storage/?p=162&tag=leftCol;post-1752

read the zdnet post with a giant helping of salt, as the math and extrapolation veers quite terribly on the wrong side, but it'll give you a good overall view (and probably a good scaring :p)

Basically, for one reason or another a drive can and will every so rarely try to read something and fail completely to read it. If it happens during a raid rebuild, your raid can crash and burn if you're running a raid 5. Generally speaking it's very unlikely to happen, but when deciding on what type of raid to go with it's important to factor everything into the decision.
 
RAID 6, with RAID 5 the only thing you gain increased performance and since you said the read/write speeds are not critical and both would be fine, i dont see a huge benifit of going RAID 5
 
+1 for RAID 6. My PERC 6/i Sees no significant performance hit from using raid 6 vs. 5.
Get around 440 MB/s sustained reads from a six disk array. That's 110 MB/s per disk accounting for parity.
Rebuilds take 5 hours. That works out to 15 hours for a 2 TB drive. (I'm using 640 gb drives.)
 
a second bit error, a bad cluster on any of the disks in the array, which will cause a read error... both are causes for second bit errors, which RAID-5 cannot protect against. additionally, the larger the drives are, and the slower their rotational speeds are, the longer a rebuild takes. this is regardless of whether it's RAID-5 or RAID-6, although RAID-6 will take longer to rebuild than RAID-5.
 
Hmmm yes I get the idea now. I don't trust these big drives much... So if a drive has bad sector this could very well happen?
If that is true, than I don't really think it is such a far fetched idea. I've seen a lot of new hdd's coming in pretty bad shapes, let alone used ones...
:)

I thought the idea of a hot spare was so there is no loss in availability or that transfer speed of the array won't be affected since hot spare is already present.
But I totally forgot that, hot spare means -> just an empty drive sitting there taking up room and power, waiting to be reconstructed in a failed raid member's image, and not meaning -> already filled drive that instantly takes over failed members place (silly me, that would mean like 6:1 chances of it being right in my case)

I don't think this R5E is for me... availability is no concern to me, I can just power down the server until i buy a replacement... in some way that is similar to me having a hotspare. Making sure I survive a 2 drive crash is what I'm after and I guess R6 is the way to go.

Losing 2 drives due to parity is also no big deal.
I filled my 6.36TB server pretty quickly, but last 500GB are going down very slowly in comparison. So I figured the next one being 10.91TB usable should add some more "security" since it seems it will be in it for a longer run. :)
 
You must log in or register to reply here.
Back
Top