Re: ZFS performance weirdness on a a Netra X1

Tim Bradshaw <tfb@xxxxxxxx> wrote:
On 2006-08-17 21:54:11 +0100, "Daniel Rock" <v200633@xxxxxxxxxxx> said:

That way ZFS cannot enable the write cache on the disks. But these
are IDE disks and on non-Sun IDE disks the write cache is normally enabled
by default. So this type of setup should cause the performance
degradation.

Yes, they are non Sun disks as you guessed. I think you mean `should
not' not `should'?

Yes, sorry for the typo.


I just did zfs create export/... with no special options.

Hmm, tomorrow I might be able to do some tests on old hardware as well
(E3500 with 400 MHz CPUs). I don't think that checksumming (default
checksum is a simple xor) should be a performance bottleneck but you
could give it a try and turn checksumming off:

zfs set checksum=off export/zones/ts


What is the load of the machine during I/O tests (top, prstat, etc.)?
How busy are the disks, what I/O pattern (iostat -xnz 5)?


--
Daniel
.

Relevant Pages

  • Re: Improvements to fsck performance in -current ...?
    ... >> Current has two major changes re speeding up fsck. ... > write cache enabled. ... Unfortunately, IDE disks do not permit disconnected writes, due ...
    (freebsd-current)
  • Re: Tips sought for optimizing 10.4.10s performance
    ... upgrade to disks with those speeds. ... Most desktop drives have been 7200 RPM, ... 7200 RPM disks have only had 2MB of cache. ... a CPU upgrade, or do I just get a new Mac. ...
    (comp.sys.mac.system)
  • Re: Tips sought for optimizing 10.4.10s performance
    ... If you're doing graphics or video editing, even more RAM. ... upgrade to disks with those speeds. ... 7200 RPM disks have only had 2MB of cache. ...
    (comp.sys.mac.system)
  • Re: Improvements to fsck performance in -current ...?
    ... >> write cache enabled. ... > Unfortunately, IDE disks do not permit disconnected writes, due ... Unfortunately SCSI drives all seem to ship with the WCE bit set, ...
    (freebsd-current)
  • RE: Write Caching
    ... If Node loses both paths to the disks I would venture to say you would see ... Your data in cache is still safe because it is held on the storage ... I have seen FTDISK events being reported even for SAN volumes. ... >> the NODES Raid Controller cards has no way to update. ...
    (microsoft.public.windows.server.clustering)