Bob Friesenhahn wrote: > On Fri, 10 Jul 2009, Roland Mainz wrote: > > - "tail" _may_ now be a bit slower since it no longer uses |mmap()| > > (which was one of the root causes for crashes (e.g. if the underlying > > file shrinks while "tail" reads it)) > > - some commands like "join" should be faster now since it uses |mmap()| > > (but we have an option to turn this behaviour off to avoid running into > > the issue described with "tail" above) > > Perhaps you did not notice the recent discussion on the zfs-discuss > list where it was revealed that I/O using mmap() on zfs can be > expected to be slower since ZFS uses an ARC rather than the Solaris > unified page cache, and therefore there needs to be extra > processing/logic to reconcile the two, and there may be double the > memory consumption as well.
I know about that problem since a long time (that's why I am not that happy about ZFS (it simply consumes too much memory and I don't have the money to simply put more memory into the machines to compensate (and half of my machines can't be upgraded with more memory anyway))) ... but for all other filesystems (UFS, QFS, VxFS, tmpfs) I'm aware of it's reverse - using |mmap()| improves performance a lot. I consider the ZFS behaviour a bug which needs to be fixed on the ZFS side, not on ours. > Once a file has been touched by mmap(), it remains slower as long as > pages referencing it remain somewhere in memory. You can easily > measure this effect for yourself. I know... "join" has the option "--nommap" for that purpose... ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.ma...@nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 3992797 (;O/ \/ \O;) _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
