ceph/doc/rados/operations/erasure-code-shec.rst

   1 ========================
   2 SHEC erasure code plugin
   3 ========================
   4
   5 The *shec* plugin encapsulates the `multiple SHEC
   6 <http://tracker.ceph.com/projects/ceph/wiki/Shingled_Erasure_Code_(SHEC)>`_
   7 library. It allows ceph to recover data more efficiently than Reed Solomon codes.
   8
   9 Create an SHEC profile
  10 ======================
  11
  12 To create a new *shec* erasure code profile::
  13
  14         ceph osd erasure-code-profile set {name} \
  15              plugin=shec \
  16              [k={data-chunks}] \
  17              [m={coding-chunks}] \
  18              [c={durability-estimator}] \
  19              [crush-root={root}] \
  20              [crush-failure-domain={bucket-type}] \
  21              [crush-device-class={device-class}] \
  22              [directory={directory}] \
  23              [--force]
  24
  25 Where:
  26
  27 ``k={data-chunks}``
  28
  29 :Description: Each object is split in **data-chunks** parts,
  30               each stored on a different OSD.
  31
  32 :Type: Integer
  33 :Required: No.
  34 :Default: 4
  35
  36 ``m={coding-chunks}``
  37
  38 :Description: Compute **coding-chunks** for each object and store them on
  39               different OSDs. The number of **coding-chunks** does not necessarily
  40               equal the number of OSDs that can be down without losing data.
  41
  42 :Type: Integer
  43 :Required: No.
  44 :Default: 3
  45
  46 ``c={durability-estimator}``
  47
  48 :Description: The number of parity chunks each of which includes each data chunk in its
  49               calculation range. The number is used as a **durability estimator**.
  50               For instance, if c=2, 2 OSDs can be down without losing data.
  51
  52 :Type: Integer
  53 :Required: No.
  54 :Default: 2
  55
  56 ``crush-root={root}``
  57
  58 :Description: The name of the crush bucket used for the first step of
  59               the ruleset. For intance **step take default**.
  60
  61 :Type: String
  62 :Required: No.
  63 :Default: default
  64
  65 ``crush-failure-domain={bucket-type}``
  66
  67 :Description: Ensure that no two chunks are in a bucket with the same
  68               failure domain. For instance, if the failure domain is
  69               **host** no two chunks will be stored on the same
  70               host. It is used to create a ruleset step such as **step
  71               chooseleaf host**.
  72
  73 :Type: String
  74 :Required: No.
  75 :Default: host
  76
  77 ``crush-device-class={device-class}``
  78
  79 :Description: Restrict placement to devices of a specific class (e.g.,
  80               ``ssd`` or ``hdd``), using the crush device class names
  81               in the CRUSH map.
  82
  83 :Type: String
  84 :Required: No.
  85 :Default:
  86
  87 ``directory={directory}``
  88
  89 :Description: Set the **directory** name from which the erasure code
  90               plugin is loaded.
  91
  92 :Type: String
  93 :Required: No.
  94 :Default: /usr/lib/ceph/erasure-code
  95
  96 ``--force``
  97
  98 :Description: Override an existing profile by the same name.
  99
 100 :Type: String
 101 :Required: No.
 102
 103 Brief description of SHEC's layouts
 104 ===================================
 105
 106 Space Efficiency
 107 ----------------
 108
 109 Space efficiency is a ratio of data chunks to all ones in a object and
 110 represented as k/(k+m).
 111 In order to improve space efficiency, you should increase k or decrease m.
 112
 113 ::
 114
 115         space efficiency of SHEC(4,3,2) = 4/(4+3) = 0.57
 116         SHEC(5,3,2) or SHEC(4,2,2) improves SHEC(4,3,2)'s space efficiency
 117
 118 Durability
 119 ----------
 120
 121 The third parameter of SHEC (=c) is a durability estimator, which approximates
 122 the number of OSDs that can be down without losing data.
 123
 124 ``durability estimator of SHEC(4,3,2) = 2``
 125
 126 Recovery Efficiency
 127 -------------------
 128
 129 Describing calculation of recovery efficiency is beyond the scope of this document,
 130 but at least increasing m without increasing c achieves improvement of recovery efficiency.
 131 (However, we must pay attention to the sacrifice of space efficiency in this case.)
 132
 133 ``SHEC(4,2,2) -> SHEC(4,3,2) : achieves improvement of recovery efficiency``
 134
 135 Erasure code profile examples
 136 =============================
 137
 138 ::
 139
 140         $ ceph osd erasure-code-profile set SHECprofile \
 141              plugin=shec \
 142              k=8 m=4 c=3 \
 143              crush-failure-domain=host
 144         $ ceph osd pool create shecpool 256 256 erasure SHECprofile