Do you use RAID?
Submitted by trollll on Wed 3 Sep 2008
| RAID 0 |
   8% | 222 votes |
| RAID 1 |
26% | 683 votes |
| RAID 2 |
0% | 8 votes |
| RAID 3 |
0% | 4 votes |
| RAID 4 |
0% | 6 votes |
| RAID 5 |
24% | 628 votes |
| RAID 6 |
3% | 93 votes |
| Nested combination |
6% | 163 votes |
| No |
29% | 770 votes |
| Total 2577 votes |
[ Parent ]
[ Parent ]
But you are right about the issues on raid-5, rebuild are very long, and problems can come fast if you not carefull.
When I loose a drives on a raid-5, I immediatly shutdown the raid and get in back online only for the rebuild, because degraded raid-5 is very bad for the drives.
[ Parent ]
I'm quite surprised by the number of votes for RAID 0. The only place I'd really consider using it is for fast, throw-away storage, like a squid cache... Other than that, I'd have thought it's a lot more niche than the numbers seem to suggest.
Cheers:wq
[ Parent ]
But what benefit do you really get using RAID0 on a workstation?
If it's a matter of having the effect of a single large block device, I'd prefer to use LVM, as it's far more flexible.
If it's a matter of speed, what are you doing on a workstation that really needs that little bit of extra disk speed, and what makes it worth the potential issues of RAID0?
Just curious really, nothing else.
Cheers.:wq
[ Parent ]
With Raid0 I almost have 80 MB/sec! In my opinion fast harddrives are better then a current CPU, plenty RAM ist good, really plenty RAM ist even better ;)
[ Parent ]
~80 MB/sec is nice, but that's just read speed isn't it? What's your write speed? My guess is that it'll be the same, or slightly worse, than a non-raid volume of the same age...
So it might be good for watching movies, but not necessarily for rendering/transcoding (home) movies.
I guess it all depends on what you use the hardware for...
Cheers.:wq
[ Parent ]
[ Parent ]
[ Parent ]
[ Parent ]
[ Parent ]
i use only raid 0 for performance
:)
[ Parent ]
On production, the database and mailserver uses a RAID50 (a striped RAID 5 array of 2 x 5 drives and both array's have a (hot)spare). Webservers use plain old RAID5.
My workstation has two (SATA2) WD Veliciraptor (15000 rpm) 150GB disks in RAID0.
We do not use RAID6 because that would require 2 parity disks (that way you can recover from two harddisk failures at the same time) and you need to really use a lot of data disks to 'justify' 2 parity disks. I mean in a 5 disk array you loose 40% of the possible space and the SCSI drives aren't cheap.
I wonder who uses RAID6 in production systems and even more why they chosen to use it.
[ Parent ]
Risk calculations (based on MTBF of 1M hours and a 2 year hardware (12 drives) cycle) set the risk that one drives fails theoretically at 0.21%.
[ Parent ]
Have you read the papers published for the 5th USENIX Conference on File and Storage Technologies (february 2007) ?
- That ones in particular:
- Disk failures in the real world: What does an MTTF of 1,000,000 hours mean to you?, Bianca Schroeder, Garth A. Gibson (Computer Science Department Carnegie Mellon University) http://www.usenix.org/events/fast07/tech/schroeder/schroeder_html /index.html Quoting the abstract:
The mean time to failure (MTTF) of those drives, as specified in their datasheets, ranges from 1,000,000 to 1,500,000 hours, suggesting a nominal annual failure rate of at most 0.88%.
We find that in the field, annual disk replacement rates typically exceed 1%, with 2-4% common and up to 13% observed on some systems. This suggests that field replacement is a fairly different process than one might predict based on datasheet MTTF
- Failure Trends in a Large Disk Drive Population, Eduardo Pinheiro, Wolf-Dietrich Weber and Luiz Andre Barroso (Google Inc.) : http://labs.google.com/papers/disk_failures.pdf
Every sysadmin concerned about raid and hard drives lifetime should read them.
regards, Xavier
[ Parent ]
Nevertheless my question was who uses RAID6 and for what reason. RAID6 can recover from 2 harddisk failures at the same time on the same array. That's the only benefit over RAID5. But for the double parity storage you loose another disk and performance is affected. What's the chance that a second disk fails before the generation of the first disk (using the hot spare) is finished?
In my book RAID6 can only be used in environments where reliability is far more important than performance and costs.
[ Parent ]
Risk calculations (based on MTBF of 1M hours and a 2 year hardware (12 drives) cycle) set the risk that one drives fails theoretically at 0.21%.
Sorry, but how do you calculate the probability that a disk fails within a year only knowing the MTBF? According to my statistics knowledge, the mean of the distribution is not enogh, you need the standard deviation as well.
[ Parent ]
[ Parent ]
As stated in my other replies, can you justify the extra costs for RAID6 over RAID5. Or; how much are you allowed to spend to minimize the risk of a second drive failure during the regeneration. We have only 1 RAID6 array and that one belongs to the administrative/financial department. All other systems (database, webserver, etc) are loadbalanced (a RAID5 with two failure can be taken out the pool without too many problems).
[ Parent ]
Very few people should need RAID 5 these days, I mean with Terabyte storage devices at under 100GBP (200USD), unless you are running a Large Hadron Collider, or trying to solve chess the hard way....
[ Parent ]
[ Parent ]
The work Linux systems use HP/Compaq hardware mirroring (RAID1). It seems to work well and when a disk dies the array keeps running and automatically rebuilds when we hot-swap in a replacement disk.
At home I've mostly only used single disk systems with back-up and never actually had a problem. When I got my newest system I want for Linux Software RAID1 and then use LVM2 on-top. In theory you can just use LVM2 and let it do the mirror, but I couldn't find a good explanation so I didn't bother.
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
[ Parent ]
[ Parent ]
[ Parent ]
[ Parent ]