• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Which RAID level?

TeeJayHoward

TeeJayHoward

Limpness Supreme
2FA
Joined
Feb 8, 2005
Messages
13,341
10 (Or 14, I've not yet decided) WD 320GB SATA (non-RE) drives
Supermicro drive enclosures with 5000RPM 120MM fans (let's hope that's adequate cooling!)
Supermicro triple-redundant 600W PSU (Included in the SC942i chassis)
2 Supermicro AOC-SAT2-MV8 controllers
8GB DVD ISOs, 100-300MB anime AVI/OGM/MKVs
XFS, Debian 3.1r1

Okay, so the data itself isn't that important. But it would be a major pain in my ass to get back. I'm looking at upgrading from 1TB to 3 or 4TB. I'd like some form of redundancy, and was wondering rather I should go with software RAID-5, RAID-5 with a hotspare, or RAID-6. With this many drives, I definately want something out there "just in case". Maybe it'll save me 30-40 hours reripping and redownloading. I'd like to have as much space as possible, obviously. Advice?

PS - I'm going with the non-RE drives because I've already got two of them. Should I mix and match instead? Or spend a bit more and buy two more RE drives?
 
I'm not an expert, but with that many drives you'd want at least a hot-spare for RAID-5. The chances of a second drive failing before you can replace the first goes up with that many. I dont know if RAID-5 w/ hotspare or RAID-6 would really make a difference in that case, wait for more knowlegable people to respond.
 
piako said:
http://www.reliasoft.com/newsletter/2Q2000/mttf.htm
Click to expand...
Interesting read. However...
TFA said:
This example illustrates the potential pitfalls of using the MTTF as the sole reliability metric.
Click to expand...
I'm looking for something closer to another user's experience with drive failure given a certain number of drives, and percentage of failure within a given time constraint. I'm less concerned with multiple failures on a single drive, and more with multiple failures spanned across multiple drives.
 
1) BUY RE DRIVES.

2) Either raid 5 + hotspare (faster, needs rebuild time if a disk is lost) or raid 6 (slower, but you can have more than 1 parity disk). I'd personally go with raid 6, but keep in mind it's gonna be pretty slow. But I could be wrong; I'd love to see benchmarks. Even just STR.

Which drive enclosures are those?

 
unhappy_mage said:
BUY RE DRIVES.
Click to expand...
Roger!
unhappy_mage said:
I'd personally go with raid 6, but keep in mind it's gonna be pretty slow. But I could be wrong; I'd love to see benchmarks. Even just STR.
Click to expand...
How much slower, would you say? Speed isn't too much of an issue, as I'll probably be the only person ever using these files, but if it'll slow it down enough that watching that 8GB ISO (Or streaming it, if I decide on a NFS setup) isn't possible, then I'd really like to know. Oh, and benchmarks will be forthcoming... Whenever my damned funds come in. I asked my school for about $3500 for a computer. All hail student loan abuse!
unhappy_mage said:
Which drive enclosures are those?
Click to expand...
Depends. I've got two CSE-M35T-1B's in my 942i, if I get 10 drives.
Otherwise, I'd be buying a Supermicro CSE-932T-R760B Black 3U Server Case, and a new motherboard and processor, and making a nifty little 3U NFS.
 
unhappy_mage said:
I'd personally go with raid 6, but keep in mind it's gonna be pretty slow. But I could be wrong; I'd love to see benchmarks. Even just STR.
Click to expand...

It'll be relatively slow, especially with writes, but looking at the data he is storing, it doesn't matter(DVD ISOs and video clips).

RAID 5 would be ok too, long as he is actively monitoring the array and prepared to get a replacement drive in a timely fashion.

Multiple drive failures where it is entirely a drive issue are pretty rare. Off the top of my head, in over 10 years of doing this stuff, I've never seen a second drive in an array fail before the first failure was replaced.* The times I've seen multiple drive failures have usually been simultaneous and due to non-drive issues---controller failure, power issue, etc.

*in an environment where people noticed the first drive failure, lol. Where nobody was paying attention, I've seen it happen twice, and of course those are the same types of environments where good backups are a misnomer.
 
RAID is about availability, not backup.

How much data are we talking about?
 
kumquat said:
RAID is about availability, not backup.
Click to expand...
Uhm, yes. I know that. *points to his original post, where he mentions the information as being noncritical and recoverable*
kumquat said:
How much data are we talking about?
Click to expand...
Well... I'd say about 10x320GB, so... 3TB or so? Maybe 4? As much as I can get. I've got more DVDs than space, and I've got about 1TB of anime on various DVD-Rs already. I could probably fill 5 or 6TB inside a month.
 
With that many drives, I would go RAID6, maybe even raid6 with a hotspare. A 10 drive raid-5 array would take quite a bit to rebuild if one of the drives died, and drives in big raid arrays can have a tendency to fail together, as the drives are usually all the same model (and possibly even batch if they are all purchased at the same time), and all of them run under very similar conditions (heat, load, etc). Would suck to have a second drive in a raid-5 array fail while the array is rebuilding, especially since its still difficult to come up with a cost-effective backup strategy for over 1TB of data... I know I'm waiting for blu-ray recorders to solve my backup needs.
 
TeeJayHoward said:
Uhm, yes. I know that. *points to his original post, where he mentions the information as being noncritical and recoverable*

Well... I'd say about 10x320GB, so... 3TB or so? Maybe 4? As much as I can get. I've got more DVDs than space, and I've got about 1TB of anime on various DVD-Rs already. I could probably fill 5 or 6TB inside a month.
Click to expand...
So what's the plan for filesystem corruption or power supply failure that takes out 8 of the drives at once?
 
Return the drives, re-rip the DVDs, download whatever he needs. This isn't important data, this array is mostly for convenience's sake (as I read it, anyways). Noncritical and recoverable are the key words.

 
RavenD said:
With that many drives, I would go RAID6, maybe even raid6 with a hotspare. A 10 drive raid-5 array would take quite a bit to rebuild if one of the drives died, and drives in big raid arrays can have a tendency to fail together, as the drives are usually all the same model (and possibly even batch if they are all purchased at the same time), and all of them run under very similar conditions (heat, load, etc). Would suck to have a second drive in a raid-5 array fail while the array is rebuilding, especially since its still difficult to come up with a cost-effective backup strategy for over 1TB of data... I know I'm waiting for blu-ray recorders to solve my backup needs.
Click to expand...
Do most people transition to new arrays once they notice disks failing often?

re: kumquat, wouldn't raid 10 work as protection against filesystem corruption?
 
Either don't use WD drives, or get RE drives.

WD non RE drives have a KNOWN issue in RAID. (thus the reason for the RE drives)

RAID 6 would be best, but you lose two drive's space worth of drive space.

Can you swing a Areca card, or is that too much money? If so, they can do RAID 6 much faster than anything else out there. They still have a performance penalty, but much less of one than doing software on the SuperMicro cards.

==>Lazn
 
Lazn_Work said:
Can you swing a Areca card, or is that too much money? If so, they can do RAID 6 much faster than anything else out there. They still have a performance penalty, but much less of one than doing software on the SuperMicro cards.
Click to expand...
Depends on the price of the Areca card, and the performance benefit it will return. Like mage said, it's just for convience sake. (Sorry for not making that clearer) If the extra $700+ for the Areca controller is the only way to read the 8GB ISOs at a decent speed, then I'll have no choice. But surely RAID-6 wouldn't reduce the performance THAT much... Would it?
 
TeeJayHoward said:
Depends on the price of the Areca card, and the performance benefit it will return. Like mage said, it's just for convience sake. (Sorry for not making that clearer) If the extra $700+ for the Areca controller is the only way to read the 8GB ISOs at a decent speed, then I'll have no choice. But surely RAID-6 wouldn't reduce the performance THAT much... Would it?
Click to expand...

Well, RAID 5 and 6 both only have write penalties. They read plenty fast.

But RAID 6 has twice the performance penalty of RAID 5:
http://searchstorage.techtarget.com/originalContent/0,289142,sid5_gci945381,00.html
"RAID-6: Dual parity. RAID-6 takes the striped parity region from RAID-5 and duplicates it. Each disk in a RAID-6 stripe has two parity regions, each calculated separately. The advantage of RAID-6 is that even if two disks fail, the RAID stripe would still have a complete set of data and be able to recover the failed disks. The biggest disadvantage is in performance. The performance impact of RAID-5 is roughly doubled as each set of parities is calculated separately. Where RAID-5 requires one extra disk for parity RAID-6 requires two, increasing the cost of the implementation."

So if you go with software, I would do RAID 5 with a hot spare.

==>Lazn
 
Lazn_Work said:
The performance impact of RAID-5 is roughly doubled as each set of parities is calculated separately.
Click to expand...
That's an oversimplification. The majority of the hit in write speed in raid 5 is when you're doing less than full-stripe writes (a single block is being updated) because you have to read from two disks before any writing happens. With full-stripe writes, XOR in software is so fast (think several GB/s) you can get raid-0-like performance out of the array. So when you do raid 6 instead of raid 5, small writes need 3 reads and 3 writes in order to get the single block updated. Full stripe writes are a little more costly, granted - raid 6 is a lot slower than raid 5 to calculate. But "a lot slower" is relative; see this from my 866 mHz p3's `dmesg`:
Code: 
raid6: sse1x2     863 MB/s
raid6: using algorithm sse1x2 (863 MB/s)
md: raid6 personality registered as nr 8
863 MB/s is nothing to sneeze at, and that's on less than a gHz of cpu. CPU power is unlikely to be the limiting factor here.

 
That's not writing anything to disk. I'm using one of the Supermicro controllers, but it's not relevant; this is just a test of the module when it loads. It sees which algorithm performs the calculations fastest. It puts dummy data through each of them to pick which will get used. On the p3, all that's available is SSE. There are SSE1, SSE2, MMX, integer-based, and Altivec implementations of the raid 6 engine. I understand the Altivec is particularly well-designed for this, but I don't have a Mac to test it with. But try "sudo modprobe raid6" and if you've got raid 6 built as a module it'll print something to dmesg; try "dmesg | tail". I bet most of your machines will beat this.

Oh, and on raid 5 it's like 1.7 GB/s. Like I said, XOR is cheap.

In real life, I get ~50 MB/s to and from disk on a half-full raid5 volume. I don't have enough free space to do any good tests right now, though :p

 
You must log in or register to reply here.
Back
Top