[OmniOS-discuss] NBD on OmniOS

[Chris Siebenmann]

| Also, the ZFS pool backing the NFS server may be slow to write (all
| NFS I/O is sync, and you may require an SSD log device to speed this
| up - seek dtrace scripts that would help you analyze beforehand if you
| have any sync IO that may be the culprit, or temporarily disable sync
| I/O on the datasets which serve your NFS roots and subjectively "feel"
| if things improve or not). [...]

 These days this is only half true unless you are still using NFS v2
(you shouldn't be, NFS v3 is better).

 NFS metadata operations such as renaming files are synchronous but
writing data is only periodically synchronous. The NFS client normally
issues a stream of asynchronous writes and then periodically asks the
NFS server to flush all pending writes (this is the NFS v3 'commit'
operation, for people using dtrace to watch NFS server activity).

 How frequently NFS clients issue NFS commit operations depends on how
fast they're writing how much (and also how they're configured; the
client must keep a copy of all uncommitted data in case the server
tells it 'oops, send that again' when the client sends in the commit
operation).

	- cks