[OmniOS-discuss] Write performance regression from r14?

[priyadarshan]

Hi Steve,

We have just completed a period of testing using OmniOS 151026, having as reference our legacy platform, FreeBSD 11.2-RELEASE.

Test was done using exact same hardware, mainly on a simple 8-disk RAID-Z2.

I/O throughput was comparable.

Did you test using r151026 as well?

I saw you opened a ticket, which is probably the favourite way.
https://github.com/omniosorg/illumos-omnios/issues/225

Priyadarshan


> On 15 Jun 2018, at 18:54, steve at linuxsuite.org wrote:
> 
> Howdy!
> 
>    I have about 20 machines built on this image circa 2015?
> 
>    OmniOS_Text_r151014.usb-dd
> 
>    I just noticed that OminOS is being continued as omniosce.org.
>    I downloaded the latest install image and did some testing,
>    and I got about 1/3 - 1/2 the write performance I was expecting on a
> simple RAIDZ setup.
> 
>    I installed fresh images  from omnios.omniti.com and omniosce.org
> based on r22
> 
>    r151022.usb-dd
> 
> as well as the r14 image above.
> 
>   I get about 1/3 - 1/2 the write performance with r22 compared to r14.
>   It is  a simple write test using dd and measuring performance with
> zpool iostat.
> 
>  Hardware and zpool is identical for each test. I simply swapped out the
> boot disks
>  and booted a different image.
> 
> The hardware is simple..  DELL R710 with LSI-SAS9201-16e
> 
>   Thoughts?
> 
>  I would like to help resolve this if it interests anyone.
> 
>  r14 suits my purpose well enough... but this issue must affect others...
> 
>     -steve
> 
> 
> 
> _______________________________________________
> OmniOS-discuss mailing list
> OmniOS-discuss at lists.omniti.com
> http://lists.omniti.com/mailman/listinfo/omnios-discuss