Redundancy
How many failures a chunk is meant to survive, and in what form file data is kept.
MooseFS keeps a chunk redundant in one of two forms. Choosing between them trades disk space against everything else. A copy is a whole duplicate of the chunk on another Chunkserver. Each copy costs the size of the chunk. Erasure Coding splits the chunk into data parts and adds parity parts computed from them, so the same tolerance costs a fraction of the chunk's size.
One figure measures both forms. The Redundancy Level counts the failures that a chunk survives while staying readable. That count makes the two forms comparable: three copies and Erasure Coding 8+2 give the same number at very different cost.
Goal is the word MooseFS used for redundancy before Storage Classes. Goal is here because you still meet the word. It appears in configuration written years ago, in older documentation, and in the names of three client tools. Today these tools do nothing but name their replacements.
Related tasks
- Choosing between copies and Erasure Coding: what each format costs in disk space, before a class is written.
- Erasure Coding starts at the ARCHIVE definition: putting in place a class that moves data out of copies.
- Chunkserver and disk recovery: what to do when the redundancy that a chunk has is not the redundancy that it should have.