Speed tests on cached filesystems don't do what is expected. The kernel
will write into RAM as much as it can, and leave the actual I/O for later.

On Thu, Jun 23, 2022 at 12:39 PM Jakub Juszczakiewicz via Dng <
dng@???> wrote:

> Hi list,
>

>      What is time between call dd? On ext filesystems is implemented
> delay before physically write date on disk. It's for e.g. minimally
> files fragmentation. Second question is that, did you try call "sync"
> command after each of writes?

>
> ---
> Best regards,
> Jakub Juszczakiewicz
> Krypto-IT
>
> W dniu 2022-06-23 14:27, Radisson via Dng napisał(a):
> > Its a normal HD i guess 4096 is ok,
> > but i do not think that this matters.
> >
> >
> > Am 23.06.22 um 13:11 schrieb Rich W:
> >> Hi,
> >>
> >> A blocksize of 1G seems extreme.
> >> What is the optimal blocksize of the output device?
> >>
> >> Cheers,
> >> -Rich
> >>
> >>
> >> On Thu, Jun 23, 2022, 4:52 AM Radisson via Dng <dng@???>
> >> wrote:
> >>
> >>> Hi list,
> >>> i found a strange ext4 (?) effect.
> >>>
> >>> when i write to a disk:
> >>> dd if=/dev/zero of=xx bs=1G count=1
> >>>
> >>> reports 2.6 GB (expected)
> >>> doing again speed drops to 200MB.
> >>>
> >>> removing xx restores old speed.
> >>>
> >>> I ask the net and it seems that the effect is there
> >>> since kernel 2.6. I found the explainations a bit confusing.
> >>>
> >>> does anyone know more ?
> >>>
> >>> note: i was analysing a performace issue for mysql and i am
> >>> not sure if that problems matters outside tests.
> >>>
> >>> re
> >>> rp
> >>>
> >>> _______________________________________________
> >>> Dng mailing list
> >>> Dng@???
> >>> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
> >>>
> >>
> > _______________________________________________
> > Dng mailing list
> > Dng@???
> > https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
> _______________________________________________
> Dng mailing list
> Dng@???
> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
>


--
Bruce Perens K6BP