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

[Gordan Bobic <gordan@xxxxxxxxxx>]

On 06/28/2014 04:57 PM, Kuba wrote:
> W dniu 2014-06-28 14:26, Gordan Bobic pisze:
>
> > No, but it does deprecate the RAID and caching parts of a controller, so
> > you might as well just use an HBA (cheaper). Covering the whole stack,
> > ZFS can also make much better use of on-disk caches (my 4TB HGSTs have
> > 64MB of RAM each. If you have 20 of them on a 4-port SATA card with a
> > 5-port multiplier on each port, that's 1280MB of cache - more than any
> > comparably priced caching controller. Being aware of FS level
> > operations, ZFS can be much cleverer about exactly when to flush what
> > data to what disk. A caching controller, in contrast, being unaware of
> > what is actually going on at file system level, cannot leverage the
> > on-disk cache for write-caching, it has to rely on it's own on-board
> > cache for write-caching, thus effectively wasting those 1280MB of disk
> > cache.
> Out of curiosity - does ZoL take control of the on-disk caches? It seems
> that (for example) FreeBSD's implementation does not:
>
> "The caveat about only giving ZFS full devices is a solarism that
> doesn't apply to FreeBSD. On Solaris write caches are disabled on drives
> if partitions are handed to ZFS. On FreeBSD this isn't the case."
>
> https://wiki.freebsd.org/ZFSTuningGuide
If there are other file systems on the disk, it disables the disk's 
write cache because there is no way to tell how it's flushing will 
interact with other file systems running independently of it. It's a 
paranoid thing to do, but ZFS is all about keeping the data safe.
There is nothing stopping you from re-enabling it using

hdparm -W1 $disk

If all of your file systems issue write barriers appropriately it should 
be safe.
If you only have ZFS on the disk, IIRC it will enable the disk's write 
cache.
Gordan

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