Re: ZFS performance weirdness on a a Netra X1

Daniel Rock <v200633@xxxxxxxxxxx> wrote:
Hmm, tomorrow I might be able to do some tests on old hardware as well
(E3500 with 400 MHz CPUs).

Ok, I have dome some tests on this ancient brick with mixed results.
It is a E3500 with 4 x 400 MHz. I have turned off 3 of its CPUs (via
psradm) to get similar results. Disks were FC attached A5200 disks
(since ZFS doesn't reset the write cache if you destroy a pool the
UFS tests did also benefit from the still enabled write cache).

I achieved with ZFS a write throughput
dd if=/dev/zero of=/pool/bigfile bs=1024k count=1024
sync
an average throughput of ~15 MB/s
But during the test the system was almost unusable (sluggish) with a kernel
time of up to 99%.

With UFS in contrast I achieved the same throughput but the system kept being
responsive during the test (kernel time at ~35%)

So the CPU is likely too slow for ZFS. Turning checksums off didn't make any
significant difference.

--
Daniel
.

Relevant Pages

  • Re: Zahl gerade oder ungerade?
    ... für aktuelle intel Core2 CPUs ... AND/OR/XOR: ... Latency: 1 Throughput 0.33 ...
    (microsoft.public.de.german.entwickler.dotnet.vb)
  • Re: Detecting CPUs and cores
    ... The reason Intel provided hyperthreading is that it increases throughput on the chip. ... core chip. ... I am not arguing that HyperThread is as effective as multiple cores. ... The trick is to discern when the number of CPUs reported by Java is a reliable number and when it isn't. ...
    (comp.lang.java.programmer)