Roch, true that a fixed-block size filesystems introduces these seek operations you talked about. now this has to be counter-balanced with "latencies" introduced in a log-structures filesystems (which personally, I am unable to list) the 10% you talked about is roughly the difference in performance between QFS and disk-raw speed, in the case of large-sequential-IO operations...in this particular cornerstone case, no other filesystem does better (not yet at least) In orther areas, a log-structure fs (ZFS) will be unbeatable what we need is to qualify, and to have a global consensus on, where zfs fits and where it doesn''t...I think we are geting there. and I am pretty sure that will end up saying something like: in "xx% of workloads, zfs is the N.1 choice" (the "general purpose" paradigm is reached if xx> 90%) s. On 3/30/07, Roch Bourbonnais <Roch.Bourbonnais at sun.com> wrote:> > Le 30 mars 07 ? 08:36, Anton B. Rang a ?crit : > > > However, even with sequential writes, a large I/O size makes a huge > > difference in throughput. Ask the QFS folks about data capture > > applications. ;-) > > > > I quantified the ''huge'' this as such > > 60MB/s and 5ms per seek means that for a FS that requires a seek per > I/O (QFS ?) > the throughput at infinite I/O size will be at most 10% better than > at 3MB I/O size. > > For ZFS the equation does not stand because it does not incur a seek > per I/O. > > -r > > > > > > (This is less true with ATA disks that tend to have less buffering > > and much less sophisticated architectures. I''m not aware of any > > dual-processor ATA drives, for instance.) > > > > > > This message posted from opensolaris.org > > _______________________________________________ > > zfs-discuss mailing list > > zfs-discuss at opensolaris.org > > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >