Windows 8 Storage Spaces...



  • So, I was intrigued by the concept behind the Windows 8 Storage Spaces, specifically the part where it allowed you to assemble a Storage Space with heterogenuous discs.

    Thus I took three empty discs of the sizes 320 GB, 1 TB and 2 TB and created a Storage Space with a Parity scheme. The manager told me that I had just created a Storage Space with a capacity of 2.0 TB. Not too shabby, I thought, considering the massive size disparities here.

    I then proceeded to fill said storage space - which led me to discover the abysmal write speed: a maximum of 20 MB/second. Okay, that can be forgiven, it is a software solution after all, though it meant that filling it to the brim would take a while (the new file copy dialog has managed to be accurate, by the way!).

    While I was letting the computer do its thing, I was suddenly greeted by the error dialogue: "Copying not possible. Destination not found." (not verbatim). Upon further inspection I discovered that the Storage Space had disappeared!

    Quick invocation of the Storage Space Manager: The storage space was still there. Just offline. Because it was full. After a mere 960 GB, mind.
    The nice thing about this manager is that he also shows the load of each individual disc: The 320 GB was at 100%. The 1 TB was at 32%. The 2 TB was at 16%. Wait a minute...

    Yeah, you're right: It filled each drive with exactly 320 GB, the capacity of the smallest drive and then decided that it was full.

    I do not think that this is how "Parity" works. Not to mention that if I had used the poor man's method of ensuring data integrity and simply copied the contents of the 320 GB + 1 TB unto the 2 TB drive, I would a) also have ensured data integrity in case of a single drive loss, b) have more capacity (1.3 TB!) and c) faster write speeds to boot.

    I think they should take a look at the load balancing algorithm, something's not right there. This is Windows 8 RTM, by the way.

    So, we have two WTFs here:

    1. A drive manager which is lying to me when it tells me about the available capacity (2 TB != 960 GB)
    2. A drive manager which, instead of displaying a warning and making the storage space read-only when full, lets it simply vanish by taking it offline.


  • Discourse touched me in a no-no place

    What you're describing sounds like a triangular version of the round wheel of RAID. Please tell me I'm wrong in thinking MS have reinvented RAID. Badly...



  • @PJH said:

    What you're describing sounds like a triangular version of the round wheel of RAID. Please tell me I'm wrong in thinking MS have reinvented RAID. Badly...

    Quite the opposite: it is BETTER than RAID. (according to Microsoft)

    @Microsoft said:

    While the RAID concepts of mirroring and striping are used within Storage Spaces, the implementation is optimized for minimized user complexity, maximized flexibility in physical disk utilization and allocation, and fast recovery from physical disk failures. Given these significant differences in objectives and implementation between Storage Spaces and traditional inflexible RAID implementations, the RAID nomenclature is not used by Storage Spaces.

    Full article



  • Well, I just discovered the next WTF: So I emptied the Storage Space and now want to delete it, to get the individual discs back.
    Then I saw this. Something does not add up here.
     



  •  Embrace, extend, excrete.



  •  Hey but don't worry. I've heard many people say Windows 8 is so great and stuff, so I'm sure it'll work out once it goes RTM. Oh... wait...



  • I think this already existed, al least in Windows Server. And it sucks great balls. We had a 16-core, 24Gb machine with a number of 320Gb disks configured as you describe, and the performance was abysmal, especially when creating small files. And guess, the task involved the creation of some sort of tiles, hundreds of millions of them. It made the 16-core machine behave like a Pentium I. Fortunately, the java app that rendered the tile, and that was written by a java performance expert, did manage to consume all of the 24Gb in a short time, so that wasn't wasted.

    The memory still makes me cringe.



  • I stumbled across another software solution named FlexRaid, though, which promises to do the same thing and more. Maybe I'll try that one out sometime in the future.



  • This issue has been known for a while. See for example this blog:

    . The take home:

    Only hard disks of identical size can efficiently be combined into a storage pool, because: (maximum size) = (size of smallest disk) * (number of disks)


  • @barfoo said:

    This issue has been known for a while. See for example this blog: . The take home:

    Only hard disks of identical size can efficiently be combined into a storage pool, because: (maximum size) = (size of smallest disk) * (number of disks)

     

    WTF? Either Windows writes parity on that (varying, coauld be none) extra area on the disks that it isn't using for data, or there is no parity data.

     



  •  Exactly what problem does Storage Spaces solve?  (purport to solve, I mean)



  • @dhromed said:

     Exactly what problem does Storage Spaces solve?  (purport to solve, I mean)

    Well, if you have a bunch of discs, you could combine them into one storage area and gain some data integrity protection at the same time.
    This would make things easier if you run out of space - when you add a new disc, you then don't have to shovel data around or find some scheme to insure that new data gets written to the new disc.

    Instead, your old partition would simply become larger. Plus, as I said, if one of the drives dies you simply replace it.



  • I'll stick with my DROBO but this RAID WTFery does remind me of a problem I had to fix a couple of years back whilst working as an IT guy for WTF University.

    Head of a research group brings me in a 'server'. This is a server only in the sense that it provides a service, its one of those "Take home user grade hardware, put it in a shiney Lian Li full tower case so it looks impressive and call it a server" jobs. Apparently it stopped booting and all his groups vital work is on there, he just needs it back up and running enough to pull the data off as they are shutting the project down soon anyway.

    The system was built by one of my co workers many moons ago, after consulting with all the IT people in his department including the guy who originally built and the fancy titled and supposedly senior "Computer Officer" the consensus was "I dunno, its fucked, just wipe it and reload it.", so he comes to me.

    The academic was understandably pissed off, especially because the original builder had specifically made him pay for double the hard disks and a mainboard with integrated RAID controller so that the whole system could be setup on mirrored disks, supposedly so that the chances of loosing anything (to drive failure) were reduced.

    I felt pretty safe guaranteeing him that I could at least get some of his data back even if I had to cleave the data out based on file headers to do it.

    So I open the thing up and the absolute first thing I notice is that the drives aren't connected to the RAID controller on the board at all, I pop them out and run cursory checks over them both and they both seem to be in physically good health however the partition type isn't recognised by the OS when I try to mount it. I take raw images including the empty space off both of them just to be safe before having a dig around inside.

    What I see looks like NTFS but the partition table doesn't match, I change the partition type on one of the drives and plug it back in, it boots. Then it becomes clear, Windows seems to think it has a software RAID with a missing disk.

    So many WTF moments:
     Building a 'server' out of cheap bits.
     Specifying a RAID controller and not using it.
     Introducing software RAID into the mix.
     The software RAID seemed to be the cause of the problem.
     Every one of the people who looked at it up until I get it basically thought that recovering data extended to plugging the drive into a USB dongle and if no window pops up, its fucked.

    I got a few bottles of wine in exchange for fixing 'the impossible' (or so he had been told by everyone else), effectively all I actually did was flip a single bit in one byte of the partition table.

     

     

    mod: man you're not even trying to format your post, are you? —dh



  • @EncoreSpod said:

    fixing 'the impossible' (or so he had been told by everyone else),
     

    This is what's wrong with our industry.

     

    – OOOH. HE'S MAGIC

    – No he's not, he's just well-informed.



  • @Rhywden said:

    Well, if you have a bunch of discs, you could combine them into one storage area and gain some data integrity protection at the same time.
    This would make things easier if you run out of space - when you add a new disc, you then don't have to shovel data around or find some scheme to insure that new data gets written to the new disc.
     

    So if you just use drives of the same size, I suppose it really is a way more friendly and usable 'home' alternative to RAID, no?

    At the same time, perhaps pros might still prefer RAID for whatever reason. I don't know much about RAID, so my comments should probably stop here. :)



  • @TGV said:

    I think this already existed, at least in Windows Server. And it sucks great balls.
    For the record, Microsoft Dynamic Disk did exist and was awesome. For striped and RAID volumes, each aspect (disk area) had to be the same size, but spanned volumes could go everywhere, it was actually reasonably performant, and it didn't have any of the problems that Storage Pools have. And now it's gone...



  •  @Rhywden said:

    I stumbled across another software solution named FlexRaid, though, which promises to do the same thing and more. Maybe I'll try that one out sometime in the future.

    Or SnapRaid. Much better suited for a) home users and b) large amout of data

     



  • @TwelveBaud said:

    ...performant..

    Internet-face-punch protocol initiated.

     


Log in to reply