On Thu, Jan 15, 2009 at 21:51, <casper....@sun.com> wrote: > > >>The performance issue of using a drive to multiple unrelated >>consumers (ZFS & UFS) is that, if both are active at the >>same time, this will defeat the I/O scheduling smarts >>implemented in ZFS. Rather than have data streaming to some >>physical location of the rust, the competition of UFS for >>I/O will cause extra head movement. > > > Solaris still makes sure that blocks are sorted, whether they > come from UFS or from ZFS. Yes, but consider the common case where UFS and ZFS are on separate slices of the disk. Then writes that ZFS thinks will be contiguous aren't, because the disk has seeked (sought?) to the UFS slice, a long way away. Solaris will optimize this as much as possible, sure, but there's nothing you can do to avoid moving the head back and forth from one slice to the other.
There's only so far you can sort blocks 12 and 10**7 ;) Will _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
