Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 16 Feb 2013 20:05:53 +0100
From:      "C. P. Ghost" <cpghost@cordula.ws>
To:        Jens Schweikhardt <schweikh@schweikhardt.net>
Cc:        questions@freebsd.org
Subject:   Re: Should I bother with a gvinum stripe when using a pair of SSDs?
Message-ID:  <CADGWnjXErEgEwYxX1BircMHkryirnD%2Bkox2=WrGMRm9cJpVvCA@mail.gmail.com>
In-Reply-To: <20130216183449.GD3070@schweikhardt.net>
References:  <20130216144751.GC3070@schweikhardt.net> <CADGWnjVJqm_s%2BWzF9_=Wt6tCg8kODbZor5-JXe-wdbiTBaZf=Q@mail.gmail.com> <20130216183449.GD3070@schweikhardt.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Feb 16, 2013 at 7:34 PM, Jens Schweikhardt
<schweikh@schweikhardt.net> wrote:
> On Sat, Feb 16, 2013 at 04:44:06PM +0100, C. P. Ghost wrote:
> # On Sat, Feb 16, 2013 at 3:47 PM, Jens Schweikhardt
> # <schweikh@schweikhardt.net> wrote:
> # > hello, world\n
> # >
> # > currently the only gvinum partition on my home system is a stripe for /home
> # > across two Velociraptor HDDs. I'm thinking of replacing the HDDs with a
> # > pair of SSDs. I was thinking of reducing complexity and in the migration
> # > possibly no longer use gvinum at all--one less thing to configure and worry
> # > about.
> # >
> # >  * Would gvinum striping bring any speed advantage with a pair of SSDs?
> # >  * Or am I hitting other limits so that striping SSDs is a waste anyway?
> # >  * Should I finally take the plunge and acquaint myself with ZFS?
> # >
> # > System has 4GB RAM in an ASUS P5Q3 Deluxe with SATA II. It appears to me
> # > that SATA II with 300MB/s is maxed out by a single SSD and striping it
> # > will not improve r/w throughput. Is my simplistic reasoning correct?
> #
> # Jens,
> #
> # as always it depends on what you're trying to achieve:
> #   - max speed / lower latency?
> #   - max storage?
> #   - max redundancy?
> #   - max run-time-to-data-loss?
> #
> # Your choice of SSD probably means you'd like to reduce latency
> # and maximize data throughput.
>
> Exactly, when I started vith vinum many years ago in the magnetic HD
> age, striping with vinum gave me almost factor 2 in r/w speed as
> measured with dd. (I do backups regularly to other media, so data
> loss protection is not my primary concern).
>
> I realize that maximum SSD speeds as advertised by vendors and tests
> (e.g. 520MB/s for contemporary top notch SSDs) may only be reached
> under certain conditions far away from my normal usage, which is
> re-building worlds and kernels and ports on a daily basis. So if for
> my realworld working set a single SSD can deliver 300MB/s, striping
> with vinum just might get me a factor 2 again to 600MB/s across
> two SSDs. Then it would be worthwile to keep gvinum.
>
> Does that make sense? My understanding of SSD and SATA capabilities
> may however be completely dreamed up...

Assuming the right cables etc.., I *guess* the limiting factor would
ultimately be the (AHCI-)SATA controller itself, or even the bus to
which it is attached. Add to this that embedded DMA controllers may
compete for the bus, limiting transfer rates even more. I don't know
how such a setup scales; that's way too system dependent. If you have
hard requirements w.r.t. latency and throughput, you'll ultimately have
to run some real world tests on the target system.

> Regards,
>
>         Jens
> --
> Jens Schweikhardt http://www.schweikhardt.net/
> SIGSIG -- signature too long (core dumped)

-cpghost.

-- 
Cordula's Web. http://www.cordula.ws/



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CADGWnjXErEgEwYxX1BircMHkryirnD%2Bkox2=WrGMRm9cJpVvCA>