SSD vs. Rotating Media

As far as I know, SSDs do not have any mechanism like DRAM to "refresh" the memory. Charges are "trapped" in potential wells, which are only slightly similar to capacitors. They are much better at holding their charge for an extended time. I only have a few data points. But, I have a laptop I generally only use for travelling, and it sits turned off under my desk for months when I'm not out somwhere. I have never had its SSD lose any data.

Jon

Reply to
Jon Elson
Loading thread data ...

You might want to be more careful when playing with glass hard disk platters: I was never able to get all the glass shards out of my office carpet after breaking a platter and was forced to replace the carpet.

IBM initially had some difficulties making reliable drives with glass platters. This was typical: However, that was 16 years ago and presumably glass platter quality has improved.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

I did some digging and found that SSD controller chips do "patrol read and patrol scrub" which reads all LBAs to repair and refresh the data periodically: However, this feature seems to be commonly disabled because it allegedly wears out the SSD with too many patrol scrub writes. I don't see how a weekly or monthly patrol read and scrub are going to make much difference over the lifetime of an SSD which is good for

10,000 to 100,000 write cycles (depending on the type of SSD): Of course, in order to do a patrol read and scrub, power would need to be applied to the SSD, and some commands sent to the SSD controller chip in order to initiate a patrol read and scrub cycle. So, applying battery power to the SSD while in storage isn't going to help here. I could not find a product that does this but did find that the Linux MegaCLI program has the necessary incantations: (scroll down to "Patrol read" section).
--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

MegaCLI is the command to control the MegaRAID raid controller boards that are commonly found in Dell servers.

These are "intelligent" controllers with an onboard CPU and the patrol read is a feature of those controllers, with similar reasoning as what you describe. The raid functionality of the array (assuming it is not raid-0) will be used to rebuild the sector when a read error occurs.

I don't think there is a 'patrol re-write' involved in this, other than that the drives itself may do that when they have difficulty reading a sector.

Reply to
Rob

I did some more reading and found that MegaCLI talks to the LSI (now Avago/Broadcom) controller chip, which as you state is on the RAID controller board. I originally thought it was on the SSD, but that's wrong. The software is at: and is available for other operating systems. Unfortunately, MegaCLI is not suitable for taking a stack of archival SSD drives, and refreshing the data on a regular schedule.

I don't know either. Patrol read and scrub can be implemented on the SSD, on a RAID/SATA/Whatever controller, or in software. My guess(tm) is that it's on the SSD. Let's pretend that data rot[1] sets in after

1 year. I write some data to some part of the SSD and then not write anything to those cells for 1 year. Without some form of data refresh, data rot will eventually corrupt the data in those cells. It would seem to me that some form of data refresh is a necessary feature. I'll do some more reading on SSD controller chips and see if it's built in, external, or missing. [1] Data rot = SILC = Stress Induced Leakage Current
--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

On Sat, 30 Apr 2016 09:03:54 -0700, Jeff Liebermann Gave us:

Especially since they are (still) all Aluminum now.

Glass platters are too expensive and the tech differences yield too little. Perpendicular head tech bested everything.

Also IBM is out of the HD business. Sold their tech to Hitachi.

Reply to
DecadentLinuxUserNumeroUno

Yes it was originally known as LSI MegaRAID but I am myself most familiar with them under the name Dell PERC.

Those RAID controllers were primarily used with harddisk drives. Originally SCSI, then SAS and SATA.

Thise drives have ECC on the disk sectors and spare areas to re-map bad sectors. A disk can re-write a sector when it had to do ECC to be able to read it, and it can re-map a sector to a different area in such cases as well. I believe control commands (SCSI/SATA commands) exist to tell the drive if it should do that.

So, by reading the entire disk regularly, one is making sure that the drive can use that option before the sector becomes totally unreadable, and in a raid controller like the MegaRAID the sector can be re-constructed from the other drive(s) in the array when it is totally unreadable, and then re-written to the failed disk.

For SSD a similar situation exists, however regular (non-enterprise) SSDs do not have CRC or ECC, so the drive will not be able to detect or correct errors. This means that any patrol reading and scrubbing has to occur well within the interval before they get bad, and it has to be done unconditionally. (with ECC one could skip the re-write step when the read occurs without issue, saving on the wear of the device. however, as you already noted, doing a scrub once a month should not affect the wear)

Reply to
Rob

On 30 Apr 2016 18:06:46 GMT, Rob Gave us:

No, silly. Re-written to a NEW, replacement disk drive.

Any time a drive starts going south, any admin worth his salt completely replaces it.

Reply to
DecadentLinuxUserNumeroUno

Completely!! Thanks. A lesser admin might try to Dremel out and JB-Weld in a new sector, huh?

Reply to
Dr. Dynamite

On Sat, 30 Apr 2016 12:09:18 -0700, "Dr. Dynamite" Gave us:

No. That is what a retarded hillbilly dumbfuck like you would suggest and likely try.

Reply to
DecadentLinuxUserNumeroUno

On Sat, 30 Apr 2016 12:09:18 -0700, "Dr. Dynamite" Gave us:

So, you are stalking me now, you stupid f*ck?

Somebody should come find you and feed you a nice lead breakfast.

Or in your case, a polonium breakfast so you can die very slowly and painfully.

That is what a retarded putz like you deserves.

Reply to
DecadentLinuxUserNumeroUno

SSD has quite a few spare blocks to assign as alternate sectors or blocks also known as overprovisioning. Eventually, the SSD will run out of spares. So, the manufacturers recommend reserving a large part of the drive for overprovisioning (which is a good thing for the manufacturer as it sells larger drives). Overprovisioning also somewhat improves performance. Since writes are block oriented, it would be easy enough to refresh the data by copying data to one of the spare blocks, erase, and write it back. Tracking which blocks have been recently refreshed is easy enough. However, that would require 4 erase cycles, which would cause some wear, but if only done monthly, should not be a problem. However, as far as I can tell, there's no such feature built into the SSD that will work if only power is applied to the SSD in storage.

As for ECC, some SSD's use R.A.I.S.E. (Redundant Array of Independent Silicon Elements): This is new to me so I'll read all about it later. Sigh, so many acronyms, so little time.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

Amazing. All of the 3.5" drives I've torn apart recently use aluminum platters. However, most of the 2.5" laptop drives seem to use glass. It's difficult to tell and I have to crack them to determine the composition. I must admit that it's been about a year since I've looked inside a 2.5" laptop drive, but I'll make a special effort on your behalf to check when I get back to the office on Monday. Remind me if I forget.

Yep. An Western Dismal bought HGST (Hitachi Global Storage Technologies) in 2012: So it is written, so it must be.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

I think the bottom line is that SSDs are not good for shelf storage of data and are not specified for that purpose. Seems to me they would be one of the more expensive forms of backup storage. So why use them? Isn't rotating media good enough for that purpose? That's what I do even though I don't use them as write once media.

--

Rick C
Reply to
rickman

On Sat, 30 Apr 2016 12:57:25 -0700, Jeff Liebermann Gave us:

Ahh... the 2.5 inch form factor. I can see where those and the smaller 1.5 and 1 inch jobs might be glass, but it still seems cost prohibitive when Al is so cheap and made quite flat with great surface quality.

So, I am sitting. Maybe I type and sit corrected... maybe.

Reply to
DecadentLinuxUserNumeroUno

On Sat, 30 Apr 2016 16:38:58 -0400, rickman Gave us:

I have mSATA drives I put various Linux distros on but I generally arrive at one I like and the others sit around unused for months on end. I have never had a problem and the Linux file system which checks out the drive after not being ran for long periods never finds a problem either.

Reply to
DecadentLinuxUserNumeroUno

For archival storage, I use various USB hard disk drives. My current collection is up to drive number 31. The last 10 or so are

1.5TB and 2.0TB Seagate's. USB hard disk drives are cheap, commonly available, cheap, fairly reliable, cheap, fast, cheap, and did I mention that they're cheap? However, I believe that they're a dying species, soon to be extinct and replaced by SSD. Time will tell.

It would be a dull world if everyone actually read the instructions and followed them.

I agree that SSD was not designed for long term archival storage. Judging by the numerous questions on the topic found by Google search, I suspect that more than a trivial number of users believe that SSD's will solve all the problems of rotating memory and are therefore suitable as an exact replacement. Therefore, it might be a good idea to see what can be done with todays SSD's, such as how long would the data last before the SSD needs to be plugged into an external refreshment machine. Think or a new product, startup, Kickstarter, dot com, etc. I'm thinking Raspberry Pi, multiple SATA port expansion board, and software to refresh or scrub several SSD drives as one time. Queue the commercial showing your data slowly rotting away while you're not looking. Yep, paranoia sells quite well.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558
Reply to
Jeff Liebermann

Replaced, maybe. Soon, I think not. SSDs have reached the point of being affordable for use in computer systems other than the lower end. It will be some time before you see them in even 90% of PCs sold. There is just too much of a difference in price. Even $10 is a lot in PCs at the lower end where most machines are sold. So what size SSD can you get for $50? Will anyone prefer that over a 1TB rotation drive that only costs the PC maker $40?

Actually, it was your post that enlightened me to the problem with using Flash for archival storage. Listen and learn.

If a user has any computer with a USB3 interface, why can't that be used with a $15 interface and PC software? I think if a commercial outfit is doing backup with SSDs they would see using this device rather cumbersome and "dangerous". I figure they would see plugging an archival drive into any system as adding an element of risk of damage. The image in a user's mind is "save it and FORGET it"... until you need it.

I expect they would rather use a different technology like rotating media in raid configurations. Otherwise, is Blueray the densest alternative? How many disks to a TB, 160?

--

Rick C
Reply to
rickman

20, if you use double-layer discs, or 40 for single-layer. Single layer are about half the $/GB, but either way it's not that different in cost from a HDD.

Cheers

Phil Hobbs

Reply to
billwether

Blu-ray is 25 GB per disk? That's pretty good! That would cost about $6 for Flash I think. So disks are roughly 10x cheaper than Flash at the moment and as you said, about the same as the cheapest rotating media. I guess rotating media wins on ease of use and density then. I can't burn a Blu-ray disk anyway. Not buying a burner just for that... maybe.

--

Rick C
Reply to
rickman

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.