View previous topic :: View next topic |
Author |
Message |
meulie l33t


Joined: 17 Jun 2003 Posts: 845 Location: a Dutchman living in Norway
|
Posted: Wed Aug 13, 2008 12:49 pm Post subject: new RAID-array: drives from different brands recommended? |
|
|
Hi all,
I'm about to purchase some drives that will form a RAID-array. Now I have read in various places that it's a bad idea to get all drives from the same brand/model/batch.
Should I really go for something like 1 Seagate, 1 Hitachi, 1 Samsung & 1 Western Digital drive? _________________ Greetz,
Evert Meulie |
|
Back to top |
|
 |
snIP3r l33t

Joined: 21 May 2004 Posts: 853 Location: germany
|
Posted: Wed Aug 13, 2008 1:29 pm Post subject: Re: new RAID-array: drives from different brands recommended |
|
|
meulie wrote: | Hi all,
I'm about to purchase some drives that will form a RAID-array. Now I have read in various places that it's a bad idea to get all drives from the same brand/model/batch.
Should I really go for something like 1 Seagate, 1 Hitachi, 1 Samsung & 1 Western Digital drive? |
hi!
what kind of raid do you want to build? hardware raid, software raid? and what raid level? i dont know the hw or sw solution will act if you use drives with different overall sector count.
i think the smallest count will be used for all other drives. my recommendation is to use identical harddrives.
HTH
snIP3r _________________ Intel i3-4130T on ASUS P9D-X
Kernel 5.15.88-gentoo SMP
-----------------------------------------------
if your problem is fixed please add something like [solved] to the topic! |
|
Back to top |
|
 |
overkll Veteran

Joined: 21 Sep 2004 Posts: 1249 Location: Austin, Texas
|
Posted: Wed Aug 13, 2008 1:32 pm Post subject: |
|
|
IIRC, it is recommended to use identical drives for RAID. |
|
Back to top |
|
 |
Cyker Veteran

Joined: 15 Jun 2006 Posts: 1746
|
Posted: Wed Aug 13, 2008 2:28 pm Post subject: |
|
|
No, the drives have to be THE SAME SIZE, but other than that it doesn't really matter.
They should be of comparable speeds, otherwise all the drives will get bottlenecked by the slowest drive, but apart from that it should be okay.
The reasoning behind the "Use different drives" thing is that multiple-drive failure will "Kill RAID dead!"
So, if you get all the same drives in one go, and it turns out there's a fault with the batch, you might be screwed (Esp. w/RAID0 which is tricky to recover data from, and RAID5, which is near-impossible to recover data from!).
By buying different drives, the chances of a mass-fault are supposedly less, but it's all down to chance really.
My personal take is that RAID stands for Redundant Array of Inexpensive Disks so Ijust buy the cheapest non-crap drives I can lay my hands on (i.e. makes like WD, Seagate, Maxtor, Hitachi, and not makes like Excel and Fujitsu ) |
|
Back to top |
|
 |
jcat Veteran


Joined: 26 May 2006 Posts: 1337
|
Posted: Wed Aug 13, 2008 9:35 pm Post subject: |
|
|
I often hear people talk about the possibility of drives from the same batch failing, and I can well believe the theory. But, the chances of more than 1 of a bad batch failing after 1000s of hours run time due to a common fault, and within a windows of say 2 days, are pretty minimal, maybe a couple within a month is possible. And in practice I've never seen this happen (I've sys-admined 100s of different servers).
Therefore, a long as you replace a drive quickly when it fails (within a day or 2), you can't go too far wrong. If you want to be really safe, then keep a redundant spare in the array for quick recovery to full redundancy level, the replace the spare when you can.
I recommend Seagate, as it's the brand I've had least trouble with over the years.
Cheers,
jcat |
|
Back to top |
|
 |
eccerr0r Watchman

Joined: 01 Jul 2004 Posts: 9932 Location: almost Mile High in the USA
|
Posted: Wed Aug 13, 2008 11:22 pm Post subject: |
|
|
I'm currently using in my software RAID5:
1 seagate 120G with 8M cache
1 maxtor 120G with 8M cache
1 maxtor 120G with 2M cache
1 post-seagate/maxtor merger maxtor 120G with 8M cache
Performance is not way too bad. The only thing to watch out for is to make sure the geometries match, or don't use the portions of the disk where they don't match. I've found the true maxtor disks are slightly bigger than the seagate-affiliated disks, so I had to use lowest common denominator and use the smallest of the set to make the array.
Personally the disks just need to be close to the same speed as possible, one lagging drive will slow down the array. I've found if I have one disk doing self tests, the whole array's performance tanks. _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
 |
meulie l33t


Joined: 17 Jun 2003 Posts: 845 Location: a Dutchman living in Norway
|
Posted: Thu Aug 14, 2008 10:02 pm Post subject: |
|
|
I am considering going for RAID10 (RAID5 would be nicer, but write hole issue has got me spooked a bit, unless a UPS can get me around that?) & using 4x Samsung SpinPoint F RAID drives (HE103UJ). _________________ Greetz,
Evert Meulie |
|
Back to top |
|
 |
drescherjm Advocate

Joined: 05 Jun 2004 Posts: 2792 Location: Pittsburgh, PA, USA
|
Posted: Fri Aug 15, 2008 4:54 pm Post subject: |
|
|
overkll wrote: | IIRC, it is recommended to use identical drives for RAID. |
I would agree with that. For performance reasons you want each drive to perform the same and the only way to achieve this is all drives must be from the same batch. Sure you can mix and match drives but that will not lead to optimal performance. With that said the difference between optimal performance and not will not be more than a few % points so if you want you can mix drives. I generally do not and I have over 100 drives in linux software raid (5 and 6) at the moment. _________________ John
My gentoo overlay
Instructons for overlay |
|
Back to top |
|
 |
drescherjm Advocate

Joined: 05 Jun 2004 Posts: 2792 Location: Pittsburgh, PA, USA
|
Posted: Fri Aug 15, 2008 5:03 pm Post subject: |
|
|
Quote: | RAID5 would be nicer, but write hole issue has got me spooked a bit |
Not sure what you mean by that. I would say there is little to worry about with corruption. If you are worrying about the array being destroyed in a power outage. At least that is not what I see on my side. To make a long story short the cpu in my home computer is defective (yet still works) so a few times a day it has memory / bus errors (machine check exceptions) but most of these are correctable. Some are not causing a kernel panic at least 1 time per week (got to swap out that cpu with the good one that sits on the shelf..) anyways none of this has caused any problems with the raid 5 array I have on that machine. And this has been going on since last year. _________________ John
My gentoo overlay
Instructons for overlay |
|
Back to top |
|
 |
jcat Veteran


Joined: 26 May 2006 Posts: 1337
|
Posted: Fri Aug 15, 2008 9:22 pm Post subject: |
|
|
Proper RAID controllers have battery backups for their write cache, that prevents any data being lost before it was committed to the drive in the event of a power failure.
However, if there is no battery backup for the write cache you can disable the write cache itself, but I guess that will lead to a slight performance hit. It's the usual balancing act between cost, performance, redundancy etc.. ..the choice is yours
Cheers,
jcat |
|
Back to top |
|
 |
drescherjm Advocate

Joined: 05 Jun 2004 Posts: 2792 Location: Pittsburgh, PA, USA
|
Posted: Fri Aug 15, 2008 9:28 pm Post subject: |
|
|
Quote: | write cache you can disable the write cache itself |
In my experience this leads to a huge performance hit. However I only have hardware raid controllers on windows since linux software raid in my opinion as good as any hardware raid on modern computers. _________________ John
My gentoo overlay
Instructons for overlay |
|
Back to top |
|
 |
meulie l33t


Joined: 17 Jun 2003 Posts: 845 Location: a Dutchman living in Norway
|
Posted: Sat Aug 16, 2008 4:15 am Post subject: |
|
|
drescherjm wrote: | Quote: | write cache you can disable the write cache itself |
In my experience this leads to a huge performance hit. |
Does anyone know of a site where these have been compared? The only pages I can find so far all refer to hardware RAID, and I am interested in the performance loss with software RAID...
Anyway... Based on this it seems to me that RAID5 in combination with a UPS that is able to shut down the server when it needs to should be good enough to get a decent system...  _________________ Greetz,
Evert Meulie |
|
Back to top |
|
 |
drescherjm Advocate

Joined: 05 Jun 2004 Posts: 2792 Location: Pittsburgh, PA, USA
|
Posted: Sat Aug 16, 2008 4:28 am Post subject: |
|
|
Quote: | Based on this it seems to me that RAID5 in combination with a UPS that is able to shut down the server when it needs to should be good enough to get a decent system... |
At work that keeps my around 15 TB of software raid 5 and 6 (and growing 750GB drives are $110 US now) up and running. I equip each and every server with its own UPS that is rated at least 3 times the total power draw of the server. This way the UPSs will stay up for 30 minutes to 1 hour before they will need to power down. The good thing is that only 1 time in the last 2 years the machines powered down outside of that they have all been up 24/7/356. _________________ John
My gentoo overlay
Instructons for overlay |
|
Back to top |
|
 |
meulie l33t


Joined: 17 Jun 2003 Posts: 845 Location: a Dutchman living in Norway
|
Posted: Sat Aug 16, 2008 5:10 am Post subject: |
|
|
Don't you mean 24/7/365.25?  _________________ Greetz,
Evert Meulie |
|
Back to top |
|
 |
jcat Veteran


Joined: 26 May 2006 Posts: 1337
|
Posted: Sat Aug 16, 2008 5:24 am Post subject: |
|
|
No need for floating point maths here. Lets just say "all year round"
Cheers,
jcat |
|
Back to top |
|
 |
drescherjm Advocate

Joined: 05 Jun 2004 Posts: 2792 Location: Pittsburgh, PA, USA
|
|
Back to top |
|
 |
|