Raid level 4
  Data striped at block (sector, cluster, track) level.
 
  Parity data kept on separate drive and done on physical sector level.
    Parity used to rebuild lost disk not to check normal read access.

  3 Disks minimum - SCSI (SATA ?)

  Large stripes and sequential access allows for processing of multiple 
    requests (files) at same time. 

  Random writes very slow because of parity calculation and store.
    Parity calculated by comparing old stripe with new stripe and 
    tweaking the difference on the parity stripe.
 
    All drives do not need to be written to.

    But drive being changed and parity drive have to be read in 
      order to calculate new parity settings and then both written.

    Write is one stripe at a time.

  No spindle synchronization.

  Excessive load on parity drive.

  Also, generally not implemented, because other schemes can provide same
    read access speed and better drive wear.