On 06/02/2010 13:18, Fajar A. Nugraha wrote:
On Sat, Feb 6, 2010 at 1:32 AM, J<jahservan...@gmail.com> wrote:
saves me hundreds on HW-based RAID controllers ^_^
... which you might need to fork over to buy additional memory or faster CPU :P
Don't get me wrong, zfs is awesome, but to do so it needs more CPU
power and RAM (and possibly SSD) compared to other filesystems. If
your main concern is cost, then some HW raid controller might be more
effective.
any real data to back your claims?
Then you need to be realistic - if ZFS consumes lets say 10-30% more CPU
but still can do several GBs (assuming your storage can handle it) on a
modern x86 box then for 99% of use cases where *much* less data is being
actually handled by a fs in real workloads the difference in CPU usage
in neglectable. This is even more so for fileservers (as in the OP case)
where the box is usually dedicated to do a fileserving only.
In real life in most environments, ZFS or not, the lvm/fs layer consume
much less than 10% of your CPU on an entry level x86 server and if ZFS
would consume a little bit more it doesn't really matter.
For example IIRC an old x4500 (older AMD CPUs) can do about 2GB/s
sustained throughput when using ZFS while still not saturating CPUs.
--
Robert Milkowski
http://milek.blogspot.com
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
