On Wed, Dec 3, 2008 at 6:37 PM, Aaron Blew <[EMAIL PROTECTED]> wrote:
> I've done some basic testing with a X4150 machine using 6 disks in a RAID 5 > and RAID Z configuration. They perform very similarly, but RAIDZ definitely > has more system overhead. In many cases this won't be a big deal, but if > you need as many CPU cycles as you can muster, hardware RAID may be your > better choice. > Some people keep stressing the point that HW raid does not include snapshots or what ever other features, or does so at cost, or ... or ... or . It seems to me like we assume that the above poster intended or implied the use of another file system on the HW raid system. The poster above did not specify a file system, so I may as well assume the comparisons is between using ZFS with JBOD vs ZFS on HW-raid. Then the features available to the administrator are essentially the same. Not the question becomes: What are the pros and cons for each? I have not tested this, but I would assume that the HW raid (forget about cheap motherboard chipset integrated "fake-raid") will save some CPU time because the raid controller has got a dedicated processor to do the stripe parity calculations. In addition the ZFS routines may have an easier time ITO selecting which disk to store the data on (only one disk to choose from). On the other hand, ZFS promises better fault detection, but presently this is temptered by several open bugs against ZFS during situations where degraded pools are present, eg pools freezing, etc. HW raid seem to have this sort of situation under control. Some HW raids may offer re-layout without losing data. ZFS does not (yet) offer this. ZFS claims better write performance in scenarios where less than a full stripe width is updated, and raid5 suffers from the "write-hole" problem. Nicely defined here: http://blog.dentarg.net/2007/1/10/raid5-write-hole ZFS updates are "atomic" - you never need to fsck the file system. ZFS will work regardless of whether or not you have a HW raid disk subsystem. So... what other benefits has ZFS got (as defined in my second paragraph) For what it is worth, have a look at my ZFS feature wishlist / AKA what it would take to make ZFS _THE_ last word in storage management: http://initialprogramload.blogspot.com/2008/07/zfs-missing-features.html _J -- Any sufficiently advanced technology is indistinguishable from magic. Arthur C. Clarke My blog: http://initialprogramload.blogspot.com
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
