Rollback RX vs TRIM - revisited

Discussion in 'backup, imaging & disk mgmt' started by TheRollbackFrog, May 2, 2015.

Thread Status:
Not open for further replies.
  1. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    Kurtis from Horizon DataSys offers the following as an explanation for Rollback RX's use of the SSD TRIM function (from Post #180), in response to my reluctance to even believe this function is really being performed by Rollback Rx...
    While I appreciate Kurtis' response in the above message, I cannot fault him for not knowing the innards of Rollback RX as well as its designers/developers do. But I surely can fault the Horizon DataSys Marketing/Develper "Party Line" (the one referenced by Kurtis above) that's offered to anyone who questions whether Rollback RX really supports TRIM... through investigation, I find the above claim to be patently FALSE. This question has nagged the Rollback user Community ever since the introduction of Windows 7 and its SSD TRIM support and the follow-on Windows 8.x... and also myself who has always found remnants of unTRIMmed SSDs while doing various testing for both HDS and other efforts.

    I decided to try and get to the bottom of this once and for all as I feel an unTRIMmed SSD will degrade at a greater rate over time than one that's properly TRIMmed... and that is not a good thing for the SSD lifespan. Possibly HDS uses a different definition for the SSD TRIM function than the technical community does, making its explanation above even slightly plausible, but I doubt it. Their developers and marketing people are clearly cognizant of what TRIM really means and that's what really bothers me about the offered explanation for how Rollback RX handles the TRIM requirement of SSD storage devices.

    Anyway, to get back to my testing (and it's actually pretty simple)... create a baseline, fill up the disk with known data patterns, take some snapshots, defrag all snapshots, delete all snapshots then uninstall Rollback RX...

    1. At first I wiped the test partition with all ZEROs (the same pattern that this SSD will TRIM to).

    2. I then created a basic W8.1 system followed by the installation of Rollback RX v10.3. The Rollback protected OS was appx. 16gB and the protected partition size was 55gB at Rollback's baseline.

    3. From both inside and outside of the LIVE Rollback protected OS I examined the unALLOCATED disk surface and found all ZEROs as expected from Step #1.

    4. I then ran a simple file generator, creating a known data pattern (different than all ZEROs) in 128mB file increments which I used to fill appx. 88% of my protected volume (about 250 files).

    5. I repeated Step #3 and found my CheckData pattern in the previously unALLOCATED disk space, again as expected from Step #4.

    6. I then took a single snapshot to LOCK in and reference all the files created above in Step #4.

    7. Repeated Step #3 and found the CheckData pattern as expected.

    8. I then created a single snapshot buffer between Step #6 and the coming deletion of all that data.

    9. I then DELETED all files created in Step #4 followed by a snapshot to LOCK in the file deletion.

    10. Repeated Step #3 and found the CheckData pattern... so the TRIM didn't occur here.

    11. Created another buffer snapshot after the file deletion lock-in (just for the heck of it).

    12. Restarted system, defragged all existing snapshots from the RBrx sub-Console, then BOOTed on into the OS.

    13. Repeated Step #3 and found the CheckData pattern... hmmmmmm, TRIM didn't occur during the Rollback defrag process.

    14. Under the Rollback UI, deleted ALL SNAPSHOTs except the Baseline and the CURRENT SYSTEM STATUS and reBOOTed into the OS.

    15. Repeated Step #3 and found the CheckData pattern which was NOT EXPECTED. At this point I would have expected Rollback to have TRIMmed all the non-referenced data from the previous file test... it did not.

    16. I then unINSTALLed Rollback to its original baseline and repeated Step #3 under both a LIVE W8.1 OS and from outside of the LIVE OS (via WinPE) and found all my CheckData pattern still resident on the SSD. Under the Rollback unINSTALLed OS, a TRIM optimization was peformed and all unALLOCATED data on the SSD reverted to its TRIM value of ZERO... TRIM was functional in the OS at the time of the test.

    Not one single TRIM operation had been performed on any of the data files created during my LIVE Rollback test above... all the data was still IN the SSD. As a result, the SSD must carry around all my CheckData (appx. 31gB) while executing its Garbage Collection cleanup functions... a massive burden. At this point the SSD believes its managing 49gB of data while the OS thinks it's only using 17.6 gB... a very large and burdensome discrepency.

    For a Rollback user, this condition can exacerbate over time as the system fills up its disk with data and "thinks" its reducing that burden by file deletions... but for the SSD the burden remains at the highest disk usage level. After a while the SSD will have exhausted its overprovisioning level (while trying to manage this non-existant data) and eventually will probably slow down quite a bit.

    Probably the thing that upsets me the most is that throughout all the TRIM discussion levels over the past few years, HDS has always told its users not to worry about TRIM because Rollback handles it just fine. The plain fact of the matter is that Rollback DOES NOT handle this just fine, and the claim that it does it patently false, as stated above.

    Feel free to ask any questions that this report may engender... I will try and expand accordingly. The BOTTOM LINE is that Rollback RX DOES NOT IN ANY WAY properly TRIM a Solid State Disk, if it even TRIMs it all all.

    I will be posting this information in the HDS Forum and asking them for a possible explanation for my findings.
     
    Last edited: May 3, 2015
  2. appster

    appster Registered Member

    Joined:
    Jun 19, 2007
    Posts:
    530
    Location:
    Paradise (Hawaii)
    Good morning Froggie, or (more likely) good afternoon to you. At present I have an HDD so I'm unaware as to how TRIM works. I'm nevertheless interested your findings as I am a Rollback Rx user and I will be configuring my next laptop will an SSD. That said I have to ask, why should Rollback Rx's be concerned with (and handle) TRIM? Shouldn't that be the responsibility of an SSD driver?
     
  3. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    Good morning, my Hawaiian friend!

    TRIM is really required by SSDs in order to manage their internal space properly. I don't want to get into the theory of how SSDs work but suffice it to say that without TRIM they become overburdened in their internal space mgmt and their "managed" space runs out (not "burned" out <overwritten> but available space to do its mgmt) which will cause the drive to slow down. TRIM is required to decrease what they call "write amplification" which is what needs to be kept as low as possible for the longest life of the SSD.

    Having said that, indeed the storage element's driver must be capable of passing the OS' TRIM command on to the SSD so it can act accordingly... and the drivers do (most of them anyway). But in the case of Rollback, the TRIM command issued by the OS is being intercepted in some way and not passed on to the storage device... this is where the problem lies. Rollback needs to do this to protect the "hidden" snapshot data on the disk, but, at some point along the way, that data does get deleted (status = no longer needed) by Rollback, and at this point in time it should be TRIMmed in the SSD. For some reason, that's not happening.

    The bottom line is that the SSD will eventually slow down due to not enough overprovisioned space available to do its job correctly, and with the WRITE APLIFICATION not being reduced to a minimum, the life of the SSD (it has a finite life) will be shortened. If it wasn't important, OSes would not feel the need to provide it and SSDs would not have the need to support it directly... they both do. For some reason, Rollback RX doesn't seem to think it's important. Or in my case... I believe the internal design of Rollback (it's quite old now) does not lend itself to allowing this function to easily be performed or managed.
     
    Last edited: May 3, 2015
  4. appster

    appster Registered Member

    Joined:
    Jun 19, 2007
    Posts:
    530
    Location:
    Paradise (Hawaii)
    Mahalo, I think I now understand this - you should post this on HDS' RB forum.
     
  5. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    I did... I don't know how long it'll stay there, though :D

    I asked them if they could provide an explanation for the results I was seeing... we'll see. If they do, I'll report back here.
     
  6. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    Not at all surprised by this finding, it will be interesting to see Kurtis's take on this,,,,,but I have a pretty good idea how he will respond. Typical HDS bull I am betting.
     
  7. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    BG, I'm not gonna guess what the problem with RBrx is in relation to TRIM... I'll await their response in the HDS Forum. I'm sure Kurtis, personally, had no specific knowledge as to how RBrx really works... he gets what information is available from HDS Software Engineering and passes it on to those of us who may have questions or misgivings about an HDS product. This will be as much a surprise to him as anyone else.
     
  8. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    Kurtis has shown no evidence that he is willing to buck the official HDS line. Even when when faced with evidence that HDSs position is wrong he either denies the validity of the evidence, gives the "I only know what I am told" line, or he ignores the posts supporting the evidence. In fact, even when he himself agrees that Rx containing drives need to be backed up to protect from problems Rx cannot deal with, he stands pat on the offical HDS line that is put forth by HDS marketing. We have seen this in the past so I do not expect much from him on this issue either. At best what will happen is probably going to be like the problems report you created,,,,,,lost in the "we will get to it when we have time because we are busy with more important matters" black hole, and likely never to be heard of again.
     
    Last edited: May 2, 2015
  9. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    No reply to your findings Froggie, either here or on the HDS Rx forum. Looks, at this point, like they are simply going to ignore your posts. Good thing these guys are Canada based or there might be a file with your name and pic on it in Washington,,,,,,along with such other well know aging undesirables (not that you are aging) as Arlo Guthrie, Joan Baez, and Bob Dylan.
     
  10. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    853
    Are HDS taking the ignore it and it goes away route?
     
  11. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    Since some of then are on the road at some MicroSloth show, they're gathering info from me and I'm sure will try and reproduce what I've seen.

    They'll most likely comment in their Forums (and maybe Kurtis will here as well) after they reproduce my test results... they need some time, I'm sure.

    Not to worry... I won't let it die :ninja:, I would genuinely like an answer.
     
  12. clubhouse1

    clubhouse1 Registered Member

    Joined:
    Sep 26, 2013
    Posts:
    533
    Location:
    UK

    Maybe Kurtis is spending his weekend off with friends and family...Perhaps, god forbid they are more important than dealing with obsessed Wilders members...Just a thought :)
     
  13. pandlouk

    pandlouk Registered Member

    Joined:
    Jul 15, 2007
    Posts:
    2,566
    Froggie, you want an answer to what?
    If RollbackRX supports trim?
    The answer is yes rollbackRX supports trim by intercepting/disabling thr trim commands before they reach the ssd.
    Or, if RollbackRX sends trim commands to the ssd?
    It does not send trim commands to the ssd, never had this ability. The fact that their users intercept their explanation as it fits them (placebo effect) is another story... For some people god forbid to question the credibility of a developer or a vendor or a company representative, anyone who does gets labeled as "obsessed" or "biased"
    Or maybe he has nothing to say/write... Just a thought :)

    Panagiotis
     
  14. Kurtis Smejkal

    Kurtis Smejkal Registered Member

    Joined:
    Mar 17, 2015
    Posts:
    253
    Location:
    Vancouver, BC
    This is a surprise to me. I'll be honest quite a few of the social media/forum guys (especially me and Sam) are in Chicago for MS Ignite. We'll be away from most
    Of our regular duties until NEXT week.

    On the upside, this conference is very educational both
    For customers and for us. I did see a gentleman today just throw the full gambit at RollBack Rx and it survived.

    If you're at Ignite please come down and say HI!
     
  15. Kurtis Smejkal

    Kurtis Smejkal Registered Member

    Joined:
    Mar 17, 2015
    Posts:
    253
    Location:
    Vancouver, BC
    Haha you are right. I'm in Chicago this week actually; but I will still be checking Wilders on occasion. If you have an issue or see a fellow user with an issue. Please ask them to email or PM me while I'm in Chicago.

    Cheers,
     
  16. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    After some thought I have decided you are 100% correct. The issue of false and hazardous advertising is a very minor issue, and the finding of Froggie, that Rx disables trim even though HDS denies this is of no real consequence. I am ashamed that I wasted Kurtis's, and Wilders members, time with my petty concerns. I stand humbled.


    Na, not really.

    I bet you would have a different outlook if you were negatively affected by these things. Good thing you are tech savvy enough to deal with these issues,,,,and I would guess you are not using SSds.
     
  17. clubhouse1

    clubhouse1 Registered Member

    Joined:
    Sep 26, 2013
    Posts:
    533
    Location:
    UK
    No, I don't... And I tend not to use software that doesn't perform as I wish or expect....Maybe I should, I was very dissatisfied with the performance of WSA (for example mods)...I might just groan on everyday. Who knows, there may be a nobel prize for the taking :)....As I have said before, the subject seems to have a limited appeal to a few core members it all seems pointless to me.
     
  18. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    Then why do you keep coming back to it? If its so boring and limited why do you bother?

    These Rx related threads are for those who have tried Rx and have had a problem or those (and this is the important group) who might be thinking of trying it. This latter group could use a warning about what they are about to possibly experience, and, if they decide to proceed, how to prevent this from happening to them.

    No platitudes sought at all.
     
  19. bgoodman4

    bgoodman4 Registered Member

    Joined:
    Jan 13, 2009
    Posts:
    3,131
    THe quote is from the OP of this thread

    No input from Kurtis about this issue, and none by anyone from HDS over on the Rx forum to Froggies posting of his findings. Thats now 9 days since the post over there and a bit longer here and HDS has not deemed it necessary to address this serious charge or even acknowledged the post itself (and Sam has been active since the posting on the Rx forum so he should have been aware of the findings outlined by Froggie). As well, no response to Mr Frogs “unOfishul” FAQ post on the HDS forum at all,,,,

    At a minimum an acknowledgment of both posts should be forthcoming from HDS, in these cases "silence is not golden".

    And you Rx users don't find this problematic?
     
    Last edited: May 11, 2015
  20. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    853
    I currently warn my clients RBRX and SSD are incompatible. While the may appear to work together, we have discovered some issues.
     
  21. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    BG, it's not forced silence... yet. With most of the HDS gang attending MicroSloth IGNITE last week, not much can be done. Sam has contacted me privately to be sure they have my system configuration correct so that they may "reproduce" this finding (it's pretty simple, really). I'm sure we'll hear from them shortly over on the HDS Forum. When we do, I'll post their findings over here under this thread as well.
     
    Last edited: May 12, 2015
  22. TheRollbackFrog

    TheRollbackFrog Registered Member

    Joined:
    Mar 1, 2011
    Posts:
    3,052
    Location:
    The Pond - USA
    Keatah, are you at liberty to share those "discovered issues?"
     
  23. Kurtis Smejkal

    Kurtis Smejkal Registered Member

    Joined:
    Mar 17, 2015
    Posts:
    253
    Location:
    Vancouver, BC
    This is correct. We were at Microsoft Ignite last week. We'll be looking at this issue in the next little while, and we'll likely post in the forum. We'll definitely be in contact about it.

    On a side note, Ignite was a lot of fun and very informative.
     
  24. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    853
    Currently I'm under NDA. In time I'll post more detail if the problems are still relevant.
     
  25. Kurtis Smejkal

    Kurtis Smejkal Registered Member

    Joined:
    Mar 17, 2015
    Posts:
    253
    Location:
    Vancouver, BC
    Who's having you sign this NDA? Just curious. If there are issues we are more than happy to look at them and address them.
     
Loading...
Thread Status:
Not open for further replies.