Re: [Xen-users] Cheap IOMMU hardware and ECC support importance

[Gordan Bobic <gordan@xxxxxxxxxx>]

On 06/28/2014 01:23 PM, Kuba wrote:
> W dniu 2014-06-28 13:25, lee pisze:
> > Kuba <kuba.0000@xxxxx> writes:
> >
> > > W dniu 2014-06-28 09:45, lee pisze:
> > >
> > > > I don't know about ZFS, though, never used that.  How much CPU overhead
> > > > is involved with that?  I don't need any more CPU overhead like comes
> > > > with software raid.
> > > ZFS offers you two things RAID controller AFAIK cannot do for you:
> > > end-to-end data checksumming and SSD caching.
> > There might be RAID controllers that can do SSD caching.  SSD caching
> > means two extra disks for the cache (or what happens when the cache disk
> > fails?), and ZFS doesn't increase the number of SAS/SATA ports you have.
> I'm not sure what happens if a read or write to the SLOG (the device
> acting as a write cache) fails. Anyone?
That's why you can have the ZIL mirrored.

> As for the reads, if a read from
> L2ARC (the read cache on SSD in our case) fails, it's just ignored
> (AFAIK) and data is read from the vdevs ("primary" storage).
Indeed.

> > How does it do the checksumming?  Read everything after it's been
> > written to verify?
> Each time data is read from the disk, it is checksummed and the checksum
> is compared with the value stored with the data. This way you know
> whether data you just read is good or not, not just that it had been
> written correctly.
Indeed, and this is one of the things that no RAID controller I'm aware 
of will do for you. A RAID card will read the data stripe and if that 
doesn't return an error from a disk, it will be assumed to be healthy. 
If you had a phantom write clobber one of the blocks in that stripe, 
your application will end up consuming corrupted data. ZFS prevents that 
from happening.


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users