Hey Edward, ZFS makes up for the performance loss due to checksumming by using a lot of smart algorithms. In some places its faster than other file systems / volume managers RAID controllers and sometimes slower. Sun is working to improve performance (of course) with some interesting stuff in the pipeline.
On 2/27/08 10:29 PM, "Edward Ned Harvey" <[EMAIL PROTECTED]> wrote: >> It is absolutely not the case that hardware RAID has across the board >> lower >> risk of data corruption, with or without caching. Caching usually >> increases >> speed, as you say. >> >> Take a look at this article if you want to be scared about data >> corruption, >> even when using hardware RAID: >> http://storagemojo.com/2007/09/19/cerns-data-corruption-research/ >> >> Theoretically, ZFS has much better data integrity than other file >> systems / >> volume managers / RAID controllers because it checksums all data and >> all >> metadata, all the time. Nothing else (that I know of) does this. > > This is really interesting. I have noticed bit errors occasionally, and I > am unusually paranoid of such a thing. It's about 50x more common in my > personal experience, with optical media than magnetic. And almost never on > a healthy ethernet. I do checksum and validate things with almost obsessive > compulsion. Especially installation & backup images. > > That being said, I didn't know ZFS checksums on the fly. I can only assume > this hurts performance, likewise if there's a lack of hardware caching, but > I also habitually use iozone to benchmark my raid systems. Typically dell > sata disks with caching raid 5 controller. > > I haven't googled for iozone results on ZFS systems, but if anyone here has > any such results, I'd be interested in comparing. > > _______________________________________________ > bblisa mailing list > [email protected] > http://www.bblisa.org/mailman/listinfo/bblisa --Peter Blog: http://pbgalvin.wordpress.com _______________________________________________ bblisa mailing list [email protected] http://www.bblisa.org/mailman/listinfo/bblisa
