[Leica] For Tina: a starting storage glossary.

[jbm at jbm.org (Jeff Moore)]

2009-08-02-15:32:07 Tina Manley:
> If it is something I've never heard of like NAS, is there a tutorial?

Here's a quick definition of terms to get you started.  I apologize in
advance if I get something wrong, because I'm just doing this from
memory instead of checking references.

  RAID: Redundant Array of Inexpensive Disks (really!).  A scheme to
  take multiple disks and cobble them together in one of various ways to
  look like a single disk, but with different size or reliability or
  performance characteristics.  Common RAID levels you might run across
  are:

    RAID-0: all the disks are stuck together to create one volume the
    size of all the disks put together.  Maximizes space and
    performance, but at the cost of reliability.  This is not what you
    want to do to keep your photos as safe as possible.  For example, a
    RAID-0 set created from two identical disks would offer
    approximately twice the space of one of the disks, potentially
    nearly twice the speed of one disk, and very approximately half the
    reliability of one of the disks.  It's what was going on inside the
    box of that 1T drive you had which was actually two 500G disks.

    RAID-1: a mirror.  Two disks (or other volumes, if you're cobbling
    together a multi-level structure of structures) which each hold
    identical data.  When you write to the array, all data is written to
    both, identically;  when you read, if all is being done right,
    if one is corrupted, somehow it's detected and you're supposed to
    get just the good data.  It's never been entirely clear to me how
    this is done in practice, but presumably it works.

    RAID-5: multiple disks in an array such that you get approximately
    the usable storage of (number of disks - 1), and if one disk dies,
    your data remains intact.  Data and checksums are cleverly spread
    around the disks so that all disks get about the same amount of
    wear.  Because of the mathematical data checksums used, I have
    confidence in the integrity of data if all is implemented well.
    You can use RAID-5 with three or more disks.
    
    RAID-6: something I've only encountered in practice recently;  it
    allows you to burn two disks' capacity in the available storage
    space in the array and gain the advantage that up to two drives may
    fail and your data will remain intact.  You're unlikely to be
    interested in this unless you array is composed of six or more
    disks.

  NAS: Network Attached Storage.  Just some pool of storage in a box
  (usually implemented as a RAID array, but not necessarily) which,
  instead of being attached directly to your computer to look like a
  disk drive, is hung somewhere on your network.  It presents volumes as
  network fileshares which your computer mounts.  With modern versions
  of the popular operating systems (Mac OS X, Linux, Windows XP or
  later) this can be a pretty convenient and reliable setup;  with
  decently fast networking (gigabit or faster wired ethernet), the
  performance can be entirely acceptable.

  The Infrant ReadyNAS boxes are sold specifically to be NAS.  The Drobo
  is sold to be direct-attached storage, but I believe they sell a
  NASsifying adapter which would allow you to hang it on the network
  instead.

  SAN: Storage Area Network.  It has many sophisticated attachment,
  provisioning and replication characteristics which I don't begin to
  understand properly.  I currently have it filed in my brain as
  something only a well-heeled corporation can afford to buy or
  maintain.