Veritas-bu

[Veritas-bu] Disk backup - Storage Hardware configurations

2006-03-24 19:23:52
Subject: [Veritas-bu] Disk backup - Storage Hardware configurations
From: tim.berger AT gmail DOT com (Tim Berger)
Date: Fri, 24 Mar 2006 14:23:52 -1000
------=_Part_2753_20399244.1143246232522
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

On 3/24/06, Sander <sander AT humilis DOT net> wrote:
>
> Tim Berger wrote (ao):
> >    Agreed.  70MB/sec writes is disappointing but acceptable for my
> >    requirements.  On a fast cpu machine you'll see better performance
> >    with software raid.  That's normal.  There's lots of other reasons t=
o
> >    go with 3ware.  Here's what led me there:
> >    - Auto-sector repair.  With big disks, this is critical.  I've lost
> >    software raids in the past due to latent bad block loss that led to
> >    more than one disk getting failed out.  All big disks have this
> >    problem.
>
> I agree with you. But Linux sw raid improves. Nowadays you can (and
> should if you care) run a background read test over the raid array.
> This detects latent bad blocks.


Detection and correction are two different things.  With sw raid, a
persistant bad block is not always recoverable resulting in failing out a
disk or hanging your system.  Hardware raids can remap them on the fly.

>    - Very reliable driver support.
>
> Dito for both Linux software raid and the disk controllers.
>
> >      smartd support.
>
> That goes for anything in Linux.


Are you using a 2.6 kernel? The smartmontools require ATA-passthrough
ioctl() calls with 2.6 for sata support which isn't there yet (unless it
came out like last week or something).

In any event, everyone should select what works best for them.

[snip]


> --
> Humilis IT Services and Solutions
> http://www.humilis.net
>



--
-Tim

------=_Part_2753_20399244.1143246232522
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

On 3/24/06, <b class=3D"gmail_sendername">Sander</b> &lt;<a href=3D"mailto:=
sander AT humilis DOT net">sander AT humilis DOT net</a>&gt; wrote:<div><span 
class=3D"gm=
ail_quote"></span><blockquote class=3D"gmail_quote" style=3D"border-left: 1=
px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"=
>
Tim Berger wrote (ao):<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;Agreed.&nbsp;&nbsp;70=
MB/sec writes is disappointing but acceptable for my<br>&gt;&nbsp;&nbsp;&nb=
sp;&nbsp;requirements.&nbsp;&nbsp;On a fast cpu machine you'll see better p=
erformance<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;with software raid.&nbsp;&nbsp;Th=
at's normal.&nbsp;&nbsp;There's lots of other reasons to
<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;go with 3ware.&nbsp;&nbsp;Here's what led m=
e there:<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;- Auto-sector repair.&nbsp;&nbsp;Wi=
th big disks, this is critical.&nbsp;&nbsp;I've lost<br>&gt;&nbsp;&nbsp;&nb=
sp;&nbsp;software raids in the past due to latent bad block loss that led t=
o<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;more than one disk getting failed out.&nbs=
p;&nbsp;All big disks have this
<br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;problem.<br><br>I agree with you. But Linux=
 sw raid improves. Nowadays you can (and<br>should if you care) run a backg=
round read test over the raid array.<br>This detects latent bad blocks.</bl=
ockquote><div><br>
Detection and correction are two different things.&nbsp; With sw raid, a pe=
rsistant bad block is not always recoverable resulting in failing out a dis=
k or hanging your system.&nbsp; Hardware raids can remap them on the fly.<b=
r></div>
<br><blockquote class=3D"gmail_quote" style=3D"border-left: 1px solid rgb(2=
04, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">&gt;&nbsp;&nb=
sp;&nbsp;&nbsp;- Very reliable driver support.<br><br>Dito for both Linux s=
oftware raid and the disk controllers.
<br><br>&gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;smartd support.<br><br>That=
 goes for anything in Linux.</blockquote><div><br>Are you using a 2.6 kerne=
l? The smartmontools require ATA-passthrough ioctl() calls with 2.6 for sat=
a support which isn't there yet (unless it came out like last week or somet=
hing).
<br></div><br>In any event, everyone should select what works best for them=
.<br><br>[snip]<br><br><blockquote class=3D"gmail_quote" style=3D"border-le=
ft: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: =
1ex;">
<br>--<br>Humilis IT Services and Solutions<br><a href=3D"http://www.humili=
s.net">http://www.humilis.net</a><br></blockquote></div><br><br clear=3D"al=
l"><br>-- <br>-Tim

------=_Part_2753_20399244.1143246232522--