Is Rollback RX safe on SSD

I would like to add the following caviat to Keatah's statement...

It's not just any currently Rollback protected sector... it's ANY DATA that was ever protected by Rollback, even if eventually made available by Rollback to Windows allocation once again.

 

Ah, semantics.. A protected sector is a protected sector regardless of whether it is in user by the OS or not.

It's amazing the amount of "precariousness" the data is in. Winding its way through 3 major maps and a layer of ECC more complex than standard HDD.

 

To be sure we're on the same page. Could you please clarify, freed up? Freed up by what and when? RBRX returning sectors to the file system? Or freed up as in RBRX being uninstalled from the system?

 

Sorry Keatah... freed up as in returning sectors to the NTFS file system allocation but the DATA space never being freed up in the SSD's block mgmt due to lack of TRIM.

Rollback does return sectors to the file system after the following conditions are met... the DATA is deleted, all snapshots referring to that DATA have been deleted AND the Rollback database has been defragmented. At this point, Windows has its allocation blocks back for use... but the SSD never gets that DATA freed up so it must carry it around (in the Garbage Collection process) ad infinitum.

 

Yes. The "freed up" sectors can never really be returned to 100% available status where TRIM can get at them because they may be potentially-in-use. This happens because TRIM is not aware of what the snapshots are doing.

One snapshot's instance of TRIM would affect another snapshot. Or in the best case you'd have snapshot 1's instance running TRIM, and then when you switch to snapshot #2, TRIM would run again based on what its map is seeing. What a cluster****! No pun intended.

Just looking at this issue from a different angle.

 

Would it work to hot/cold RAW Backup with IFW the rollback rx enabled drive.
Uninstall rollback and reboot
let trim do it's thing
then raw cold restore the rollback enabled drive?

That way you will still have your snapshots, but the drive will have been trimmed.

 

Jolly, that will not work. Remember, your backups are ALL SECTOR backups (both used and unused sectors)... you will get all the unTRIMmed data as well as any Windows thinks is in the current image. Even if you TRIM the drive, you will be restoring all the previously TRIMmed data from your ALL SECTOR image.

 

Thanks, Froggie. Just a thought. I really hope the HDS geniuses manage to figure out a way round that problem and the Windows update issue.

 

If HDS concentrates on the problem, reproduces it and FIXes it... all should be fine.

As far as updates go, with W7 & 8 you're kinda stuck with their Update Clearance Thread located on their Forum... and as far as W10 HOME is concerned, I have no idea how they're gonna deal with that... and come July 29th, that problem's gonna get real serious.

 

another question. If I protect C and D drives. Each protected by rollback rx and each a physical SSD Drive. Can I hot backup and cold restore just C drive and still retain rollback protection for both drives.

 

Jolly, I've not tested this personally but if you properly HOT backup with IFW and select both partitions (C: & D on the SSD, you should be able to COLD restore C: only and all should be well.

 

Where are the redirection maps stored on a multi-disk installation? What sort of information must come from "C" in order to operate "D" correctly?

I have not tested this either. I would tend to think the answer is no. I would guess the snapshot allocation maps are stored on the boot drive. If not, there's probably some sort of tie-in with snapshot management and the startup console. Restoring the boot drive would be altering how "D" is seen too.

50/50 either way.

I am fairly certain, however, that this theory can be tested on HDD and the results would be the same as on a double SSD.

 

Jolly, Keatah's point should be taken very seriously, I was waaaaay too casual with my answer above. We just don't know where RBrx keeps its database info and snap diskmaps nor do we know the form of those. If C: contains all protection info than you can be in trouble returning a C: image to a system at a later time. If each partition contains its own info, all should be well.

I would make no assumptions about this configuration at all... it needs to be very carefully tested (and protected if you're gonna test on the same SSD/HDD... FULL disk images before testing).

Edit: if I get a chance I'll pull out one of my ol' laptops and give the above configuration a whirl.

Edit2: Jolly, this might be a better question for our "resident" HDS Senior Account Representative who should be able to get an answer for you. Kurtis... can you help?

 

Results were inconclusive at best. I took 4-snaps with an ALL SECTOR IFW image after snap 2. Restored the image and the C: partition was at snap 2 and the D: partition was at snap 4, datawise... Rollback showed 2-snaps. I then Rolled back to snap 2 and when done, both the C: and D: partition were at snap 2. Also experienced an "Improper Shutdown" from the Rollback sub-Console and a "ChkDsk" w/no errors following the BOOT after the restoration.

I'm not really sure what this means, that's why I used the word inconclusive. For now I would ALL SECTOR image all protected partitions until you hear back from Kurtis/HDS on whether that type of operation is safe or not.

Personally I've NEVER protected more than my System partition whenever I used Rollback... it just always felt funny to me (putting too much at risk, I think)

 

This is a very good example of the Facts according to Kurtis and HDS.

 

Again the facts according to Kurtis,,,,which have been shown to be ,,,,,,, well not too factual.

 

Well that's certainly good to hear for a change! Go HDS! It's good to know companies can turn around.

 

From what I have been reading on the web lately, it does indeed seem debatable if an SSD without TRIM will have a shorter lifespan. Modern SSDs have advanced garbage collection and/or DuraWrite technology. This still makes TRIM desirable if you have it, but in no way indispensable. All Apple third party SSDs and most RAID configurations do not have TRIM, and there have not been any reports of dying SSDs on such systems.


Cheers
manolito

 

Do you know what DuraWrite technology is?

 

Some kind of Data Reduction method developed by LSI (Sandforce)
http://www.techspot.com/news/52835-...-need-for-trim-overprovisioning-and-more.html

 

Well actually it is data compression. Less data written = less wear on the NAND cells.

 

Panagiotis, that's very true for an early point in time while running under Rollback RX. The problem lies in the fact that as Rollback, via its user, eventually releases unneeded snapshots, and those snapshots are eventually defragged under Rollback and returned to the Windows allocation scheme for future use, all the data used by those snapshots continues to be carried around by the SSD and included in any Garbage Collection operations (it was never TRIMmed). Eventually (and pretty quickly if any large snapshots, temporary or otherwise get played with) all the free and over provisioned space gets used up and the poor SSD is left to only its READ/ERASE/MODIFY/WRITE operations to manage the space. Once that starts, ERASE/WRITE operations are now spewed across the entire SSD finding room for the new data and WRITE AMPLIFICATION is increased by a huge amount. That "plenty empty space for over provisioning" mentioned above can disappear very quickly. And this phenomenon can be intensified by RANDOM WRITE operations to the device such as in the use of extensive databases... that's probably the worst of conditions.

As I mentioned earlier, with just a few (2-3) large data snapshot operations, I can chew up an entire SSD in a very short time (that's what I had to do during the DEMO with HDS to demonstrate non-TRIMming), causing it to enter the mode described above.

 

Overprovisioning is not a temporary buffer area outside of the stated capacity of the drive. It is a limited consumable. Typically between 5-10 percent of capacity. Once it is gone it's gone.

 

Hi

Is there a way to handle this SSD problem - i am thinking like totally uninstall RollbackRX now and then and let the trim do its thing before reinstalling and begin totally newly all over again . Would that work ?

 

Well, that's the only way to "handle" it currently. It's up to HDS if they want to rewrite RBRX to properly use an SSD.