• 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.

How exactly does RAID-1 work?

J

Jaxel

Weaksauce
Joined
Feb 8, 2007
Messages
119
I am using this MOBO:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813131833

I want to put in two Samsung 840 128gb SSD drives in a RAID-1 array. This obviously is for data-redundancy. I want it so that if one of the drives failed, I am able to continue my work without being inconvenienced. My question of course, is how does this work exactly?

Both drives working in RAID-1 works fine. But what happens when one of them fails? Will the RAID simply display a warning for 5 seconds or something, telling me one of the drives is dead, and then continue on as a single drive? Or will I have to disable the RAID to continue?

After then, when I replace the drive, will the RAID automatically copy the contents of the working old drive to the new drive?
 
when I tested raid1 using software windows raid I was able to pull a drive and the system carried on as if nothing happened, there was a just a balloon popup near the system tray alerting me a drive failed but that was it. Onboard raid's may possibly react different tho. Generally even if a system were to crash I would expect it to still boot and work ok on the surviviing disk.
 
I run a RAID 1 of HDDs set up through the chipset with Intel Rapid Storage Technology installed. While neither of my drives have failed, I have had situations where I have to force a shutdown, which sometimes causes one of the drives to either lose sync or become corrupted (idk what is actually happening on a logical level).

During bootup, between POST and Windows startup, there is a screen where your drives and their health statuses are listed. In the case of my mobo, the status is usually NORMAL. With the irregular shutdowns, sometimes a drive will load up as DEGRADED (iirc, that's the word it shows), but the system continues to start up, presumably off the working drive. Once logged into Windows and Intel RST starts up, it will start a repair process in the background. In about 1 or 2 hours, the 300GB array is back to normal status.

I know I didn't answer your question but thought it just might be helpful. I'm very curious myself what sticking in a new drive into a failed array would look like to the end user. If the RAID is created somehow with serial numbers used as identifiers to map the SATA port configuration, then a new dive may not work...I really have no clue.

On a side note, you probably don't want to use RAID 1 for SSDs as TRIM is not supported for RAID 1. Especially because you got the Samsung 840 (in a small size) and not the 840 Pro, you definitely want to avoid situations where your RAID has to rebuild (I'm assuming the rebuilding process makes a hefty amount of writes to NAND). The 840 is the first consumer device using TLC NAND so I would be as conservative as possible when using it.
 
rastaban said:
On a side note, you probably don't want to use RAID 1 for SSDs as TRIM is not supported for RAID 1. Especially because you got the Samsung 840 (in a small size) and not the 840 Pro, you definitely want to avoid situations where your RAID has to rebuild (I'm assuming the rebuilding process makes a hefty amount of writes to NAND). The 840 is the first consumer device using TLC NAND so I would be as conservative as possible when using it.
Click to expand...

The 840s I got are MLC, not TLC:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820147192

What would be the downside of not having TRIM in a RAID-1 array?
 
Jaxel said:
The 840s I got are MLC, not TLC:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820147192

What would be the downside of not having TRIM in a RAID-1 array?
Click to expand...

I see. You said 840 not 840 Pro in the OP, so I figured TLC.

Anyways, it's actually not that big of a deal. Having TRIM would restore your drives from a dirty state to like-new performance state faster than the built in garbage collection algorithm on the 840 Pro's controller can on it's own. However, Samsung's GC is fairly aggressive and is able to restore performance over time on its own as shown in AT's review. Looking at that first chart under the Performance Over Time & TRIM section, it looks like it took about 1 hour and 14 minutes to restore performance after a torture test.
 
It really depends on the controller/software. One of the main things you just want to ensure is that if a drive does fail, you get an alert of some kind otherwise it's completely pointless. So when you first set it up, power down one drive to see what happens.

I've heard stories of people losing entire SANs because enough drives failed. They had no alerting setup so they never realized the drives failed till the entire thing went down. They also had no backups. :eek:
 
The traditional RAID1 scheme will never work absolutely perfect with SSDs. Even if you would find an implementation that correctly passes TRIM, as soon as you need to rebuild one of the SSDs the RAID layer would have to know which sectors or chunks actually contain relevant data to make an exact 1:1 mirror also conatining the unused flash blocks. Since this information does only exist within the filesystem, the RAID layer usually makes a 1:1 copy of the data of remaining disk and writes zeroes to the unused blocks. Of course you could make a filesystem-wide TRIM after that, but it would still reduce the number of write-erase cycles of these blocks by one.

The current implementation of linux software raid has write-intent bitmaps and can reduce the rebuild to those sectors that have been written to, but I would guess that this will not work with drives that were not in the raid before.

The best solution would be filesystem-integrated RAID like it already exists for ZFS and BTRFS, which only have to replicate block with actual data.
 
Red Squirrel said:
I've heard stories of people losing entire SANs because enough drives failed. They had no alerting setup so they never realized the drives failed till the entire thing went down. They also had no backups. :eek:
Click to expand...

283259.jpg
 
omniscence said:
The traditional RAID1 scheme will never work absolutely perfect with SSDs. Even if you would find an implementation that correctly passes TRIM, as soon as you need to rebuild one of the SSDs the RAID layer would have to know which sectors or chunks actually contain relevant data to make an exact 1:1 mirror also conatining the unused flash blocks. Since this information does only exist within the filesystem, the RAID layer usually makes a 1:1 copy of the data of remaining disk and writes zeroes to the unused blocks. Of course you could make a filesystem-wide TRIM after that, but it would still reduce the number of write-erase cycles of these blocks by one.

The current implementation of linux software raid has write-intent bitmaps and can reduce the rebuild to those sectors that have been written to, but I would guess that this will not work with drives that were not in the raid before.

The best solution would be filesystem-integrated RAID like it already exists for ZFS and BTRFS, which only have to replicate block with actual data.
Click to expand...

But there is no need for a physical 1:1 equivalence at all. On HDDs it's very important for performance issues, on SSDs not really.
 
Just the raid1 mirror would need to be made smarter to make that happen, the same way people work with sparse files.

a trimmed sector is suppost to return all 0's, so if you read a sector that is all 0's, you can assume it was trimmed, and if it wasn't, no harm, as you get the same result anyway.

Same method is used to resparse sparse files.

I recently had to recover a raid10, and added this into my rebuilding program, when it unstripped the data, if a block was 0's, write that out as sparse, so I ended up only needing 36gigs, instead of 1tb for the resulting reconstructed data. (I have a habit of zero'ing free space every now and again). Really speed up the process a lot, as that was tons of data I didn't need to write.
 
So do you guys think it would be okay for me to make a RAID1 out of the SSDs I selected?

Or should I just get some normal hard drives? If so, what hard drives would you guys recommend exactly? Small capacity is fine; I don't need more than 128gb... and I'm looking for high speed drives.
 
Aesma said:
But there is no need for a physical 1:1 equivalence at all. On HDDs it's very important for performance issues, on SSDs not really.
Click to expand...

With the rebuilt array not being a 1:1 copy I meant that you end up with one drive (the old drive) with a lot of "unallocated" (trimmed) sectors and one drive (the new drive) being completely allocated (untrimmed) unless it's overprovisioned. The second drive will degrade much faster in performance.

But this an academic problem, since I suppose the OP uses Windows and there is no trim-capable RAID1 available there. (Or does the Windows software mirroring support it?) Hence I would recommend overprovisioning both drives before creating the RAID array. I would not consider the intrinsic 7% overprovisioned space of the mentioned drives to be enough.
 
I honestly I have no idea what you guys are talking about... and yes, I am using Windows 7
 
Jaxel said:
I honestly I have no idea what you guys are talking about... and yes, I am using Windows 7
Click to expand...

LOL!

Any 2 drives can be made into a RAID1 BUT RAID1 (or any RAID) is not considered a back-up.

Just buy one SSD and connect a SATA drive (regular mechanical drive) either in an external case or mount it into the computer BUT only connect the drives when you back-up your "C" drive (everyday drive).

If that doesn't "compute" hopefully someone else will come along to explain it further.
 
I'm not doing it for back-up reasons, for that we use Carbonite. I'm doing this for ZERO DOWNTIME reasons. If a regular drive is connected, and fails, yes, we have backups, but there is downtime related to restorting backups. My hope is that with a RAID1, we can just disconnect the failed drive and continue on; then replace the failed drive when we get a chance. Zero downtime.

So going back to what I said before: So do you guys think it would be okay for me to make a RAID1 out of the SSDs I selected?

Or should I just get some normal hard drives? If so, what hard drives would you guys recommend exactly? Small capacity is fine; I don't need more than 128gb... and I'm looking for high speed drives.

I am using Windows 7 and plan on using the onboard raid controller with the motherboard I mentioned in the OP.
 
I wish you would have explained this earlier. :)

Without getting all technical, your plan sounds just fine.
 
RAID 2 of these HDDs: http://www.newegg.com/Product/Product.aspx?Item=N82E16822236245&name=Internal-Hard-Drives

But it really sucks that Intel STILL only has two SATA 6gbps ports off the chipset because an SSD on SRT is required to take up one of those ports, forcing any RAID setup onto controllers off the Intel chipset. Whether or not there is a noticeable performance hit, I don't know, but in your case where off-chipset ports are all 3gbps, burst reads from the HDD's will surely be capped.

So in the end, until Intel increases the number of native SATA 6gbps ports, there is no ideal solution for HDD RAID + SSD on SRT reliability and performance.
 
raid1 isnt a backup as such I agree, but I dont consider backing up to a seperate hdd in the same machine superior to raid1. On a raid1, all data on the raided drive is automatically a duplicated copy on the 2nd drive, whilst doing backups you have to remember or have a good automated system to backup everything you need. In addition normal backups have downtime if something goes wrong whilst a raid1 if good should carry on running without downtime if a drive fails.

As far as I know the latest 11.5 intel bios and drivers allow TRIM on raid. But you need the bios as well not just the drivers.
 
rastaban said:
On a side note, you probably don't want to use RAID 1 for SSDs as TRIM is not supported for RAID 1. Especially because you got the Samsung 840 (in a small size) and not the 840 Pro, you definitely want to avoid situations where your RAID has to rebuild (I'm assuming the rebuilding process makes a hefty amount of writes to NAND). The 840 is the first consumer device using TLC NAND so I would be as conservative as possible when using it.
Click to expand...

on a bigger sidenote, STOP TELLING PEOPLE THIS!!!!

go look at http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm there are samsung 830s that haven't died that are over 800tb of constant writes. you are not going to force your ssds out of write cycles, raid1 repairs generally only happen after a drive fails, and trim isn't going to save your drive from nand exhaustion.

quit it!!! you're just confusing users who are trying to setup their home solutions. the garbage collection is fine.
 
As the original OP is using Win7 and is unfamiliar with RAID, here's what I would suggest if he want a simple RAID1 mirror his boot SSD's. First off, use "Windows" built-in mirroring capability within the Disk Management utility. Why? Because modern CPU's have plenty of power to support it and secondly, TRIM "works". There wont be any significant difference in performance over using the Intel RAID features. depending on the chipset SATA controller OROM and driver version, the Intel RAID "might" support trim, but the built-in Win7 mirroring definitely does and is a lot easier to implement regardless of chipset. It's easier as in simply get windows up and running, start Disk Manager, tell it to mirror the boot drive with the 2nd SSD.

Another benefit of of a Win7 built-in mirror is that you basically have two bootable OS disks that work anywhere. For example, my kid's systems stopped botting up because of drive corruption. I yanked a Win7 mirror RAID1 SSD out of my system, popped it into the kid's PC, booted up, and after a few automatic driver changes, was able to use utilities on my drive to fix the drive in the kid's PC. Once I was done, I simply hot swapped it back into my system and used Disk Manager to rebuild the mirror.

The biggest weakness of the built-in mirroring is that there are no notification tools built into windows. Basically, if a drive fails, there is no auto notification built in without actively opening up Disk Manager and seeing the RAID is healthy. There are some manual ways to set up the event viewer to toggle a system event notification in case of RAID1 mirror failure, but nothing else.

So for a cheap simple software RAID1 using Win7 Built-in RAID1 capability:
Pros:
1. Easy to implement
2. Trim works without issue on RAID1 mirrored SSD's setup with Disk Manager (one of the biggest pro's when using SSD's for this type of simple RAID1)
3. Performs as well as the Intel chipset RAID1
4. One of mirrors can be yanked at any time and is capable of being used as a boot drive in most any other machine for bootup emergencies and rebuild is just as easy.

Cons"
1. No notification if a disk fails in the RAID; you have to check it once in a while or setup pain-in-the-ass event code notifications.
 
Jaxel said:
I am using this MOBO:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813131833

I want to put in two Samsung 840 128gb SSD drives in a RAID-1 array. This obviously is for data-redundancy. I want it so that if one of the drives failed, I am able to continue my work without being inconvenienced. My question of course, is how does this work exactly?

Both drives working in RAID-1 works fine. But what happens when one of them fails? Will the RAID simply display a warning for 5 seconds or something, telling me one of the drives is dead, and then continue on as a single drive? Or will I have to disable the RAID to continue?

After then, when I replace the drive, will the RAID automatically copy the contents of the working old drive to the new drive?
Click to expand...

Will the RAID simply display a warning for 5 seconds or something, telling me one of the drives is dead, and then continue on as a single drive? Yes, the warning depends on the controller. For ICH9R I get a warning at boot and in Windows

Or will I have to disable the RAID to continue? No don't do that.

Get a new drive and it should auto rebuild when you replace the drive.
 
gjs278 said:
quit it!!! you're just confusing users who are trying to setup their home solutions. the garbage collection is fine.
Click to expand...

i'll confuse them further. mirroring with SSDs in a home environment is a moronic waste of money.

with good quality SSDs, like the ones the OP mentions, you're SSD isn't going to randomly die at some point in its life. it will die when the NAND begins to wear out which will be roughly the exact same time the other identical drive in the mirror will also begin failing.
 
I disagree. SSDs fail. It can be the controller, it can be a bug in a firmware (even the best ones had critical bugs), it can be a power surge or whatever. If by home environment you mean your 6yo kid computer, then sure, but if you'd rather spend some money (that is not ridiculously high like 2-3 years ago) on a SSD RAID1 so that you don't have to wait an RMAed drive to recover your home computer then I don't see the problem.
 
madrebel said:
i'll confuse them further. mirroring with SSDs in a home environment is a moronic waste of money.

with good quality SSDs, like the ones the OP mentions, you're SSD isn't going to randomly die at some point in its life. it will die when the NAND begins to wear out which will be roughly the exact same time the other identical drive in the mirror will also begin failing.
Click to expand...

Yep exactly. Mirror raid is pointless with SSDs unless you purposely wear one out first so they wont fail at the same time, but considering it takes LOT of writes to wear one out you'd probably have to be writing to it continuously for a day so you can make a big enough offset.

While the possibility of a random failure is there (controller etc) over time I think they will become reliable to a point where it's a non issue. Better off to use SSDs strictly for the OS, image it, and have a spare one on hand. You will get downtime if it fails, but with raid1 you probably would too as they'd both fail close to around the same time.

Though I guess with raid 1 you could also monitor the SMART data which tells you how much life it has left, then change one at a time when it his 98% or something. As an OS drive it's probably good for like 10+ years anyway.
 
I think you have guys have an aweful lot of faith in manufacturing and quality control processes etc. if you think every ssd is going to fail at the exact same time.

Its not unknown in electrical products for products with same model number's to have different components as they only have to achieve the marked features. Even with same components failures can occur at different points.

Also eg. what if someone gets a ssd, runs it standalone for some months then decides to add another to create a mirror? then there is some months in the timeline gap.
 
Regarding Intel RST RAID 1 itself, I had a customer drop off a system last week that had developed a vibration. It turned out that one of his two Hitachi UltraStar 1 TB drives in a RAID 1 configuration had developed the vibration. The drive wasn't showing any verification errors but I pulled it from the system to prevent any possible corruption of the good drive. All I had to do was enter the Intel RAID setup at boot, break the mirror, let it boot in to Windows, shut it down, pull the bad drive and then boot back in to Windows to make sure everything worked properly. I had the customer take the system home to use while waiting for the replacement drive. His replacement arrived yesterday and all I have to do is stop by his place later today, pop the drive in, boot into Windows and let RST build the new mirror.

Easy as can be.
 
You must log in or register to reply here.
Back
Top