back to article Hunt the RAID killer: Exablox builds object storage that isn't

Exablox was cofounded by a video-making, rock-climbing CTO intent on killing RAID. El Reg got a closer look at Exablox by attending a press tour at its HQ last week and looking deeper into its object storage-based filer. Co-founder and CTO Tad Hunt explained that he became frustrated with filer performance while taking time …

  1. Destroy All Monsters Silver badge
    Paris Hilton

    Buzz Overload!

    RAID rebuilds take too long and disk failure likelihood during a rebuild increases with rebuild time. That pointed Exablox towards object technology but with filer access.

    Thus:

    The OneBlox filer involves customers buying an empty 8-slot Oneblox node with 4 x 1GbitE ports, at $10K list, and populating it with 3.5-inch disk drives which the Exablox OS turns into a filer resource using an underlying object storage technology. All the management is cloud-based and the system discovers new nodes and adds them into a clustering scheme, the OneBlox Ring, using peer-to-peer technology.

    I see the "not a RAID thing" but I don't see the "disk failure likelihood during a rebuild does not increase with rebuild time." If your disk croaks, new copies of the objects previously on that disk need to be created from the objects elsewhere on the system. A subsequent disk croak will still propel your data into never-never land.

    1. Captain TickTock

      Re: Buzz Overload!

      It says the opposite: "[RAID]disk failure likelihood during a rebuild increases with rebuild time"

      If I understand correctly, more advanced filesystems like ZFS (and I suppose this "SmashFS") are much quicker to rebuild, because they know to only rebuild the data which was on the dead drive (and not the whole drive). And so they stress the remaining drives less.

      Which begs the question - how is OneBlox better than the likes of ZFS and BTRFS, etc? (ZFS is designed to scale out to Zettabyte scale, hence the name)

      1. SeanExablox

        Re: Buzz Overload!

        ZFS and BTRFS are scale-up file systems predicated on RAID Volumes, Groups, and LUNs.

        OneBlox is a scale-out object-based file system and, you're correct, therefore we know exactly which objects are in any given location (drive/OneBlox/geography) at any given point in time.

        Re how we differ: ZFS has fixed controller resources and as you add capacity there's a point where the application is over utilizing the resources. (note: many ZFS systems can perform well). With OneBlox as you granularly scale, you're scaling capacity, performance, and availability as the number of OneBlox increase in the Ring--all within a single global pool of storage.

    2. SeanExablox

      Re: Buzz Overload!

      We use 3-way replication to protect against 2 drive or 2 OneBlox failure. You're correct we only need to copy the objects from the failed drive therefore regardless of the capacity of the drive it's a function of the number of objects that dictates the time back to full health (and the larger the Ring, the shorter amount of time is required). Additionally, all drives in the Ring are available to bring the system back to health so there's no need to allocate any drive as a global hot spare (as with RAID).

  2. Anonymous Coward
    Anonymous Coward

    Have I not seen this before?

    It look very similar to the IBM XIV and HP 3PAR type technologies with the "RAID" performed on blocks/chunks/containers or whatever name you would like to use.

    1. SeanExablox

      Re: Have I not seen this before?

      IBM and 3Par have done good things to mask the limitations of RAID, but are still tied to them (at the lowest level). With Exablox, we break every file into object(s) and manage them without any RAID. We employ an object replication strategy to protect information within a given OneBlox, Ring, or across geographic distances.

  3. Hetz Ben Hamo

    Out of curiosity ..

    Sean,

    Which OS do you use?

    Do you plan to open source some parts of it?

    1. SeanExablox

      Re: Out of curiosity ..

      @Hetz, we use linux at the user space layer. We have no plans to open source our IP, it's one of our major differentiators vs. the industry.

  4. Trevor_Pott Gold badge

    Exablox

    From what I've seen of the units, they seem promising (though admittedly one has not crossed my lab as yet.) One thing I will say, however, is that I like whoever did Exablox's industrial design. For a filer, they're really pretty. It's silly, but they always make me smile because of it.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2022