ceph/doc/dev/osd_internals/erasure_coding.rst

   1 ==============================
   2 Erasure Coded Placement Groups
   3 ==============================
   4
   5 Glossary
   6 --------
   7
   8 *chunk*
   9    When the encoding function is called, it returns chunks of the same
  10    size as each other. There are two kinds of chunks: (1) *data
  11    chunks*, which can be concatenated to reconstruct the original
  12    object, and (2) *coding chunks*, which can be used to rebuild a
  13    lost chunk.
  14
  15 *chunk rank*
  16    The index of a chunk, as determined by the encoding function. The
  17    rank of the first chunk is 0, the rank of the second chunk is 1,
  18    and so on.
  19
  20 *K*
  21    The number of data chunks into which an object is divided. For
  22    example, if *K* = 2, then a 10KB object is divided into two objects
  23    of 5KB each.
  24
  25 *M*
  26    The number of coding chunks computed by the encoding function. *M*
  27    is equal to the number of OSDs that can be missing from the cluster
  28    without the cluster suffering data loss. For example, if there are
  29    two coding chunks, then two OSDs can be missing without data loss.
  30
  31 *N*
  32    The number of data chunks plus the number of coding chunks: that
  33    is, *K* + *M*.
  34
  35 *rate*
  36    The proportion of the total chunks containing useful information:
  37    that is, *K* divided by *N*. For example, suppose that *K* = 9 and
  38    *M* = 3. This would mean that *N* = 12 (because *K* + *M* = 9 + 3).
  39    Therefore, the *rate* (*K* / *N*) would be 9 / 12 = 0.75. In other
  40    words, 75% of the chunks would contain useful information.
  41
  42 *shard* (also called *strip*)
  43    An ordered sequence of chunks of the same rank from the same object. For a
  44    given placement group, each OSD contains shards of the same rank. In the
  45    special case in which an object is encoded with only one call to the
  46    encoding function, the term *chunk* may be used instead of *shard* because
  47    the shard is made of a single chunk. The chunks in a shard are ordered
  48    according to the rank of the stripe (see *stripe* below) they belong to.
  49
  50
  51 *stripe*
  52    If an object is so large that encoding it requires more than one
  53    call to the encoding function, each of these calls creates a set of
  54    chunks called a *stripe*.
  55
  56 The definitions are illustrated as follows (PG stands for placement group):
  57 ::
  58
  59                  OSD 40                       OSD 33
  60        +-------------------------+ +-------------------------+
  61        |      shard 0 - PG 10    | |      shard 1 - PG 10    |
  62        |+------ object O -------+| |+------ object O -------+|
  63        ||+---------------------+|| ||+---------------------+||
  64  stripe|||    chunk  0         ||| |||    chunk  1         ||| ...
  65    0   |||    stripe 0         ||| |||    stripe 0         |||
  66        ||+---------------------+|| ||+---------------------+||
  67        ||+---------------------+|| ||+---------------------+||
  68  stripe|||    chunk  0         ||| |||    chunk  1         ||| ...
  69    1   |||    stripe 1         ||| |||    stripe 1         |||
  70        ||+---------------------+|| ||+---------------------+||
  71        ||+---------------------+|| ||+---------------------+||
  72  stripe|||    chunk  0         ||| |||    chunk  1         ||| ...
  73    2   |||    stripe 2         ||| |||    stripe 2         |||
  74        ||+---------------------+|| ||+---------------------+||
  75        |+-----------------------+| |+-----------------------+|
  76        |         ...             | |         ...             |
  77        +-------------------------+ +-------------------------+
  78
  79 Table of contents
  80 -----------------
  81
  82 .. toctree::
  83    :maxdepth: 1
  84
  85    Developer notes <erasure_coding/developer_notes>
  86    Jerasure plugin <erasure_coding/jerasure>
  87    High level design document <erasure_coding/ecbackend>