back to article Shingled drives get SpectraLogic archive down to 9 cents/GB

Spectra Verde DPE is a Shingled Magnetic Recording (SMR)-based NAS disk platform for high-density bulk storage at costs as low as 9 cents/GB raw. Shingled drives have sets of overlapping write tracks with side-by-side narrower read tracks to increase density. The downside is that a whole set of write tracks have to be …

  1. Lee D Silver badge

    All very nice and all, but with prices starting at $34k and going up to $682k, it's likely out of the range of most people reading this, I'm guessing? If you're spending that much money, probably raw size isn't a huge worry.

    To then have software RAID on that kind of hardware seems a little... out of place?

    1. John H Woods Silver badge

      "

      " ... it's likely out of the range of most people reading this ..." -- Lee D

      Quite possibly, but unlikely to be out of the range of all of the clients that many of us are advising ;-)

    2. Alan Brown Silver badge

      "To then have software RAID on that kind of hardware seems a little... out of place?"

      It's more flexible and generally faster/more reliable than hardware raid controllers.

  2. Anonymous Coward
    Anonymous Coward

    If you're spending that much money, maximum raw size for minimum power/space usage is a primary concern.

    This achieves impressive drive density - for comparison, the Backblaze storage pod gets only 45 drives in 4U. What matters though is how well it is cooled, and changing one of the drives in the middle of such a deep box could be painful.

    Why do you think software RAID is inherently worse than hardware RAID? When a drive fails, a hardware RAID would be forced to scan all 8TB across all drives; it could take days to rebuild. ZFS would only rebuild those parts of the filesystem which have files in them, and gives you strong checksum data protection too.

    1. Bronek Kozicki Silver badge

      I guess those thinking software RAID worse than hardware are stuck in the 90s, when overhead from software RAID (e.g. as implemented in Windows NT) was having an impact on writes in the operating system. In this case we have dedicated storage appliance, where by definition all hardware resources are for one purpose only, i.e. running ZFS filesystem. Using software RAID (rather than "hardware" one) makes it easier to scale these resources, since it is basically running on more or less generic hardware platform.

  3. Henry 8

    Independent failures?

    The statement "probability of data loss to 1 in over 2 million years when properly monitored and maintained" sounds rather fishy to me. What are the chances that they've taken "failure rate for a single disk" and just multiplied that up N times, assuming that all failures are independent? Even if you declare lightning strikes and earthquakes as outside the calculation, disks that are hosted in the same environment, and which were probably all made in the same production run, don't have independent failure rates.

    1. James 100

      Re: Independent failures?

      Presumably the calculation is that each RAID-Z3 array will spend N hours each year degraded/rebuilding, then calculated how many years it would take before a further double-disk failure occurred during one of those windows of vulnerability, to get a nice impressive number. Nice bit of benchmarketing, but not remotely realistic of course.

      (I preferred the guy - from Intel, IIRC - going through disk reliability figures, pointing out that with a modern drive capacity approaching 10^14 bits, suddenly that "good" error rate of 10^-15 becomes a virtual guarantee that you *will* hit corruption at some point while rebuilding your RAID 5 array - so you need a minimum of RAID 6 just to be able to rebuild an array of large drives reliably. Meanwhile, a scary proportion of our data isn't checksummed, so when it does get corrupted we won't know about it until it's too late!)

      1. Tom Samplonius

        Re: Independent failures?

        "...that "good" error rate of 10^-15 becomes a virtual guarantee that you *will* hit corruption at some point while rebuilding your RAID 5 array..."

        And the SpectraLogic product has triple parity, so issue is also solved. Plus, a weekly data scrub is important on all NASesl. And for an archive NAS with shingled disks, read IO should be quite cheap.

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 © 1998–2019