Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 23 Jan 2010 11:19:33 -0500
From:      jhell <jhell@DataIX.net>
To:        Artem Belevich <fbsdlist@src.cx>
Cc:        Alexander Leidinger <Alexander@leidinger.net>, FreeBSD Hackers <freebsd-hackers@freebsd.org>
Subject:   Re: Setting "zfs_arc_max" value in FreeBSD 8.
Message-ID:  <alpine.BSF.2.00.1001231114110.18316@pragry.qngnvk.ybpny>
In-Reply-To: <ed91d4a81001222119y3ee56181ofc66813c987620a@mail.gmail.com>
References:  <7f14551c1001190119x46c6b04dx2362cd1252f0d81@mail.gmail.com> <hj3v0i$i2p$1@ger.gmane.org> <7f14551c1001190216w49814186n1ada2b721380502b@mail.gmail.com> <4B55C5A6.2020109@DataIX.net> <20100120111433.25801pnmhrxnirok@webmail.leidinger.net> <ed91d4a81001220847v15e26830la2d51479c3d104e1@mail.gmail.com> <alpine.BSF.2.00.1001222123400.3696@pragry.qngnvk.ybpny> <ed91d4a81001222119y3ee56181ofc66813c987620a@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help

On Sat, 23 Jan 2010 00:19, fbsdlist@ wrote:
> Good to hear that it's usable for you even on a relatively low-memory
> system. Now, throw in an SSD for L2ARC, more RAM for ARC (and L2ARC
> housekeeping) and then it starts to really shine.
>

Can not wait until I can get something like this spinning!. I'm looking at 
around 6 months before this type of configuration will come into effect 
and be viable. I'm excited!.

> As for better than expected performance, in my not-so scientific
> benchmarks (copying 10G-large files on 8-disk RAIDZ2 pool until
> filesystem is full)  ZFS on FreeBSD did beat the hell out of
> OpenSolaris on the same hardware. I was really surprised. I'm sure
> something needed to be tuned on OpenSolaris, but it was nice to see
> FreeBSD performing so well.
>
> --Artem
>

These guys really are doing one hell (maybe 2 or 3 ;]) of a job. Couldn't 
ask for better people to work on this.

>
>
> On Fri, Jan 22, 2010 at 6:39 PM, jhell <jhell@dataix.net> wrote:
>>
>> On Fri, 22 Jan 2010 11:47, fbsdlist@ wrote:
>>>>>
>>>>> Anyone know if it is adjustable on a system with 1024MB of ram ? Is this
>>>>> just being auto calculated by some other value ?
>>>
>>> You may want to make sure that vm.kmem_size is set to a value much
>>> larger than vfs.zfs.arc_max. Default value may be too small to allow
>>> such a large ARC.
>>>
>>> On a side note, I'm not sure that ZFS is a good match for system with
>>> only 1G of RAM. By trial and error on my box with 8G or memory I've
>>> figured out that I need to set arc_max ~1G below physical memory size
>>> to avoid lockups under load. YMMV.
>>>
>>
>> ZFS on this box with 1G has been quite enjoyable actually. With the settings
>> I have posted I have not had any lockup on stable/7 and no sudden freezes or
>> waits for transfers. So this entirely thus far has been a godsend. I had
>> even put this thing through some of the tortures that others have posted to
>> the list and not come up with the same results but better. There is
>> obviously a lot of variables in this between hardware and configurations
>> used so the results are minimal in comparison. With ZFS in place on this
>> machine it performs a little bit under specs for the hardware but I wouldn't
>> expect anything less for such a file-system.
>>
>> --
>>
>>  Thoughts & Prayers out to Haiti.
>>
>>  jhell
>>
>>
> _______________________________________________
> freebsd-hackers@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe@freebsd.org"
>

-- 

  jhell




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.1001231114110.18316>