On 01/15/10 12:59, Jeffry Molanus wrote:
Sometimes people get confused about the ZIL and separate logs. For
sizing purposes,
the ZIL is a write-only workload. Data which is written to the ZIL is
later asynchronously
written to the pool when the txg is committed.
Right; the tgx needs time to transfer the ZIL.
I think you misunderstand the function of the ZIL. It's not a journal,
and doesn't get transferred to the pool as of a txg. It's only ever written
except
after a crash it's read to do replay. See:
http://blogs.sun.com/perrin/entry/the_lumberjack
The ZFS write performance for this configuration should consistently
be greater than 80 IOPS. We've seen measurements in the 600 write
IOPS range. Why? Because ZFS writes tend to be contiguous. Also,
with the SATA disk write cache enabled, bursts of writes are handled
quite nicely.
-- richard
Is there a method to determine this value before pool configuration ? Some sort
of rule of thumb? It would be sad when you configure the pool and have to
reconfigure later one because you discover the pool can't handle the tgx
commits from SSD to disk fast enough. In other words; with Y as expected load
you would require a minimal of X mirror devs or X raid-z vdevs in order to have
a pool with enough bandwith/IO to flush the ZIL without stalling the system.
Jeffry
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
