3 ms·
Just thinking out loud: does doing copy on write result in less filesystem fragmentation (assuming the data is not held in RAM)?
by RexRollman 15y ago
Just thinking out loud: does doing copy on write result in less filesystem fragmentation (assuming the data is not held in RAM)?
- hashbo 15y agoIn theory yes, as it avoids holes in the databases or files used for storage. It’s also the semantics of ZFS - http://en.wikipedia.org/wiki/ZFS http://en.wikipedia.org/wiki/ZFS
- ora600 15y ago... and ZFS copy-on-write has been known to cause severe performance degradation for databases. Not to mention that too many SAs keep the default ZFS block size, which is far too large for OLTP databases.