General > General Technical Chat

modern storage is crazy

<< < (13/14) > >>

tooki:

--- Quote from: jpanhalt on April 29, 2022, 06:28:01 pm ---Pure dogma and semantics.  What happens when backups are destroyed in a fire or flood?  Same scenario.  There are degrees of safety offered by redundancy.  Backups are just another form of redundancy.  As reliability of storage increases, the need for multiple methods of redundancy decreases.
--- End quote ---
No, it's not just "dogma and semantics".

RAID protects against precisely one thing, and nothing more: drive failure.

It doesn't protect against any other form of data loss. It doesn't protect against natural disasters, theft, software failure, or human error.

The key difference between backups and redundant storage is that in the latter, the redundancy data is always updated simultaneously with the original. (In the case of mirroring, the mirrored copy is updated simultaneously.) It's redundancy of your normal working data.

Backups, on the other hand, are separate copies made periodically, but not in real-time. In proper backup systems, versioning is also implemented, so that rolling back to a prior version is possible.

tooki:

--- Quote from: james_s on April 30, 2022, 12:22:16 am ---If drives had gotten more reliable the need for backups is decreased. I don't think drive reliability has changed substantially over the last 30 years or so though. Nothing is ever perfect, backups of critical data are always good to have, although there is a broad range of data that is technically replaceable but inconvenient, so one has to weigh the effort of keeping meticulous backups against the effort of reconstructing if something fails.

--- End quote ---
Nonsense. Drive failure is but one of a multitude of reasons why backups are needed. I've been lucky and have never lost data to drive failure (I replace a drive as soon as any of the critical SMART parameters start to rise, even if the disk has not yet thrown any disk errors, let alone failed). But I have used backups many times to recover things I edited and then wanted an old version of, or from program crashes or bugs.

tooki:

--- Quote from: amyk on April 30, 2022, 02:24:59 am ---
--- Quote from: tooki on April 29, 2022, 06:10:16 pm ---
--- Quote from: eugene on April 28, 2022, 01:45:23 am ---I suppose the OP is looking for some 4MB memory cards too.  :-DD

Just kidding around; I totally get his frustration. Ten years ago I bought a Synology NAS and stuffed two 2TB drives into it (RAID 1.) I thought 2 TB was ridiculous at the time and I was right; the drive is still only half full. The really sad thing is that the drives are 5.25". If one of them fails I'm not sure I could even find a replacement!

--- End quote ---
You definitely did not buy 5.25” hard disk drives for your NAS ten years ago, since they stopped making 5.25” hard disks over 22 years ago (and even then they were extraordinarily rare). Not even old stock, since the last one of those was just 19GB.
--- End quote ---
They made 5.25" disks up to 47GB:

[video]

--- End quote ---
Sure, but the last 5.25" drives made were the Quantum Bigfoots, which topped out at 19GB, according to the interwebs.


--- Quote from: amyk on April 30, 2022, 02:24:59 am ---If drive manufacturers were interested in using 5.25" platters again, 47TB would probably fit in that size. The drive above has 14 platters.

--- End quote ---
By the 1990s it was clear that there was no interest in doing that, and then not long after started going the other way, towards 2.5" platters for server and enthusiast disks.

Additionally, the economies of scale of 3.5" drives meant that merging a bunch of those into a RAID array is cheaper than the equivalent storage might have cost in a low-volume giant server disk. But the RAID is not only cheaper, but also far faster, and can add redundancy.

I also would expect that the manufacturing yields on large platters are disproportionately higher, for the same reason that physically large ICs have lower yields, because the chances of having a bad region on a big disk are higher, for a given defect rate per surface area.

DiTBho:

--- Quote from: mariush on April 30, 2022, 11:24:40 am ---4TB WD Red Plus drive

--- End quote ---

I am reading that Seagate put a kind of protective layer on the top of the magnetic platters in their 7200.xx Barracuda lines, but without testing the reliability of the solution, and as result, the protective layer can seriously damage the tiny heads, so one day you experiment the BSY-bug, which is not only a firmware problem because behind it there is that quality of your hard-disk platters sucks.

Built and tested for decent quality rather than built, poorly tested, but promoted to be great, crazily fast and cheap!

tooki:

--- Quote from: DiTBho on April 30, 2022, 11:15:56 am ---I don't want a fast disk, I want a reliable disk!

--- End quote ---
Well, you can buy some high-reliability enterprise disks (some types do have orders of magnitude better MTBF specs than others!), but for most things, you're better off with an array or SSD.

Plus a backup. Well, two backups if you wanna do things right: one at home and one off-site. (I keep my offsite backups in my locker at work and bring them home every 2 or 3 weeks to update them.)

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod