Main Area and Open Discussion > Living Room
Should we pre-emptively retire old hard drives?
Stoic Joker:
So, just buy 2 then? :) Or 4 for mirrored raid?-Renegade (August 03, 2012, 01:30 PM)
--- End quote ---
:huh: ...A 4 drive (RAID1) mirror? :D
SeraphimLabs:
MTBF is based on time between any two failures.
Most drives have an initial burst of failures when new, perhaps as much as half of the drive's failures for it's first year of service take place within the first few hours of operation.
But what happens on the SSDs is like conventional drives they on-the-fly reallocate bad sectors. Although they claim an MTBF on par with a conventional drive, the controller is less tolerant. After so many such failures have taken place, the controller panics because it is out of space to remap into.
A conventional drive will keep right on ticking, marking the bad spots and simply responding with a progressively lower capacity until something mechanical fails or the controller burns out. On the other hand the current SSDs tend to just keel over when they reach that condition, being unable to cope with so many failed areas on the media.
Though they are improving, and are quite consistent in their failure rates- as shown by multiple drives under similar conditions failing in quick succession, they still need to make them die gracefully instead of abruptly giving up.
40hz:
So, just buy 2 then? :) Or 4 for mirrored raid?-Renegade (August 03, 2012, 01:30 PM)
--- End quote ---
:huh: ...A 4 drive (RAID1) mirror? :D
-Stoic Joker (August 03, 2012, 02:26 PM)
--- End quote ---
Umm...Maybe Renegade is thinking of a RAID-10 (aka: "striped mirrors") since he's talking four disks? That would give the best performance and reliability since you get the benefits of both striping and mirroring. RAID-10 is used in many mission critical setups, although newer RAID controllers can give it a run for its money on performance using just RAID-5. It does rebuild more quickly than RAID-5 does following a drive replacement however. (I've only seen one in-production RAID-10 in my entire career BTW.)
The really big downside for personal use is that a controller that supports RAID-10 will be on the expensive side - and tends to be very fussy about the drives it uses. This is definitely more of an 'enterprise' level solution.
8).
superboyac:
So, just buy 2 then? :) Or 4 for mirrored raid?-Renegade (August 03, 2012, 01:30 PM)
--- End quote ---
:huh: ...A 4 drive (RAID1) mirror? :D
-Stoic Joker (August 03, 2012, 02:26 PM)
--- End quote ---
Umm...Maybe Renegade is thinking of a RAID-10 (aka: "striped mirrors") since he's talking four disks? That would give the best performance and reliability since you get the benefits of both striping and mirroring. RAID-10 is used in many mission critical setups, although newer RAID controllers can give it a run for its money on performance using just RAID-5. It does rebuild more quickly than RAID-5 does following a drive replacement however. (I've only seen one in-production RAID-10 in my entire career BTW.)
The really big downside for personal use is that a controller that supports RAID-10 will be on the expensive side - and tends to be very fussy about the drives it uses. This is definitely more of an 'enterprise' level solution.
8).
-40hz (August 03, 2012, 04:32 PM)
--- End quote ---
This is the big struggle I'm in right now. Given the situation of a personal computer geek which is (like me) we have a bunch of hard drives we've collected over the years. Different sizes, different manufacturers, etc. How can we stick all these in a box and divy them up into various pools? Like 3 drives in one pool of random sizes, another 3 to back that one up, a couple more for dumping stuff into, etc.
But I have a couple of other threads about this, sorry....
40hz:
I don't think there's a single anything currently off the shelf that runs on Windows wirkstation that will do all that. But you could mix & match tools to get it. Windows 8 will be including drive pooling and RAID-like features if that's any consolation. Or so the plan says.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version