• 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.

ssd/hdd raid 5?

E

ep0x73

2[H]4U
2FA
Joined
Sep 5, 2013
Messages
2,663
This might sound dumb but two SSD's in raid 0 with a HDD for parity, raid 5

The raid 0 portion would prolong read/write cells with the parity disk for redundancy.

Sound doable or would the HDD slow down the works?

Generally as I understand it parity once calculated should not factor in that often unless new files are introduced.
 
The raid 0 portion would prolong read/write cells with the parity disk for redundancy.
Click to expand...
raid5 does not work that way. It stripes parity over all disks in the array.

http://en.wikipedia.org/wiki/Standard_RAID_levels#RAID_5

Sound doable or would the HDD slow down the works?
Click to expand...
Writes will be at hard drive speed. Reads would be random but probably much closer to hard drive performance than ssd on average.

You can do what you want with SnapRAID however not in real time and its better for large files that do not change often.
 
Last edited:
Not sure what I was thinking......spaced off parity is spread across all 3+ disks.

I suppose if you want to run a SSD with some type of backup the rapid restore might work but you still probably cannot have TRIM.
Least you could manually backup when you want and not have it auto.

Course if you go that route just as easy to have windows create a backup and update it manually too.
 
Your best options would be RAID1 with two SSDs (no TRIM) if you really need the redundancy or automated incremental backup of the SSD to a large mechanical drive.
 
Still confused on some aspects. Some SSD's [maybe all?] have built in garbage collection like 840 evo, etc.
That said I also read without trim you lose about 10% overall bandwidth/speed.

Raid 1 as of now has no pass through thus no trim but if the SSD's themselves can do cell cleanup maybe it's not such a impact.

Raid5 with 3 SSD's would be interesting to say the least but expensive to implement.
Should be able to hit reads that are over 1,000 though.
 
RAID3 does what you are asking, no idea about performance though, never done a RAID3, only had one controller that supported it
 
omniscence said:
Your best options would be RAID1 with two SSDs (no TRIM) if you really need the redundancy or automated incremental backup of the SSD to a large mechanical drive.
Click to expand...

what you mean with "RAID1 with two SSDs (no TRIM)" ?
TRIM is disabled when two SSDs are in RAID1 config?
 
vFX said:
what you mean with "RAID1 with two SSDs (no TRIM)" ?
TRIM is disabled when two SSDs are in RAID1 config?
Click to expand...

Yes. As far as I know the Intel driver does only support TRIM for RAID0. But I never tried it for RAID1 or found conclusive information about that so I am not completely sure.

It is also not trivial to implement for a very simply reason:
If the OS issues a TRIM for a certain block of the RAID1, the RAID driver has to forward it to both drives.

TRIM is by the specification a command that the SSDs can do with what they want. They are free to ignore it, which is not a bad decision if the block is not aligned or smaller than the SSD page size or even a larger chunk of data. SSD vendors may chose not to map the smallest possible unit (a sector) in their mapping tables, but it is the smalles possible unit that can be trimmed. So now it can happen that one SSD ignores the command while the other SSD clears the block. Even worse, some SSDs read random data for a TRIMmed block, so even if both of those SSD execute the command, they would no longer contain equal data.

All this would not be a problem for the filesystem, since it makes no assumptions of the contents of a cleared block. But for the RAID layer this can be devastating, since it assumes the same contents on both disks. On the next data check, the RAID layer will detect a lot of inconsitencies and try to correct those, resulting in unneeded writes to the disks.

While all this may still work in principle, it is much easier to just disable TRIM. There are some SSDs which have a deterministic TRIM that results in a zeroed block, which would qualify those drives for a RAID1 with TRIM, but the reasoning to the user why some SSDs work and some don't are maybe too much of a hassle. For example, the 840 Pro guarantees zeros, the 830 and 840 EVO do not. That are the drives I could test.

This is also the likely reason why the vendors of RAID controllers do not implement that feature. And enterprise SSDs are built to properly work without TRIM (or the SAS enquivalent).

The most elegant solution from my point of view is the integration of the RAID layer in the filesystem like ZFS and btrfs do. Those filesystems could easily do TRIMs on SSDs, and btrfs already does as far as I know.
 
Last edited:
Good read, we might not see RST have built in raid function for 1, 5, 10 until 14, maybe even 15.
Then again if it was possible since they do support 0 maybe it's just not possible.
Drives with their own garbage collection should essentially keep themselves clean in a matter of speaking.

RTT might be a better solution to keep a bootable copy of your OS. Once the mirror is done and set manually just keep up with it once a week instead of on the fly.
Sure trim is also disabled as it's a hybrid raid1.
You could use a same sized HDD though for the backup copy instead of another SSD, cutting cost but not losing performance.
 
Last edited:
Most SSDs don't need TRIM if they are overprovisioned. And if they don't have enough internally you have to add a bit yourself. My 256 GB SSDs run usually at 240 GB or 200 GB depending on workload if I use them without TRIM.
 
To get max speeds with ssd its raid-10, anything else is slower than a single drive without raid
 
drescherjm said:
raid3 or raid4 are very rare.
Click to expand...
All Areca cards support RAID3. You won't get the desired performance though. Maybe if you were using an SSD for the parity disk and had a bunch of hard drives it would make a difference compared to RAID5 with all hard drives, but seeing as you have multiple SSDs, won't do you much good.

The closest thing to what you might want is Intel's matrix RAID, but it still doesn't quite fit the bill.
 
You must log in or register to reply here.
Back
Top