[prev in list] [next in list] [prev in thread] [next in thread] 

List:       evms-devel
Subject:    RE: Re: [Evms-devel] Feature request: RAID4 without stripping (only
From:       "James Lee" <james_lee31 () hotmail ! com>
Date:       2006-02-15 20:24:17
Message-ID: BAY105-W315DCE18AE1E66FDEEC708EFA0 () phx ! gbl
[Download RAW message or body]

Yeah, i also think this would be very good functionality to add to the Linux md \
driver (and subsequently to EVMS).  
For more information on this, there's a pretty huge thread over at avsforum about \
this: http://www.avsforum.com/avs-vb/showthread.php?t=573986.  One of the main guys \
running the company that markets this (limetech) has posted a lot to that thread.  I \
posted to it a while back asking if there were any plans GPL the driver code, and \
he's said that it's something they're planning to do that (releasing the driver code, \
but not the management interface).  See the following post in the thread: \
http://www.avsforum.com/avs-vb/showthread.php?p=6500421#post6500421.  
I don't know what's come of the plans to release this code, but he seems like a \
helpful chap, so it might be worth getting in contact with him.  I don't know if Neil \
Brown would be interested in adding this functionality to the kernel?



> Subject: Re: [Evms-devel] Feature request: RAID4 without stripping (only parity)> \
> To: cartman@medsite.info> CC: evms-devel@lists.sourceforge.net> From: \
> steved@us.ibm.com> Date: Wed, 15 Feb 2006 09:34:50 -0600> > Konstantinos Skarlatos \
> wrote on 02/15/2006 07:18:24> AM:> >> -----BEGIN PGP SIGNED MESSAGE----->> Hash: \
> SHA1>>>> Unraid is different from RAID4, because it does no striping. It is just>> \
> a parity drive, that is added in a system of indepependent disks of>> varying \
> capacity and file systems. The data on the data drives are not>> touched by the \
> whole operation, you could add a full drive on the system>> and the only thing that \
> would happen is the recalculation of parity on>> the parity drive. I think Talik \
> described it well:>> [quote]>> Traditional>> RAID 4>> 01 02 03 p>> 04 05 06 p>> 07 \
> 08 09 p>> 10 11 12 p>>>> Traditional>> RAID 5>> 01 02 03 p>> 04 05 p 06>> 07 p 08 \
> 09>> p 10 11 12>>>> UNRAID>> 01 05 09 p>> 02 06 10 p>> 03 07 11 p>> 04 08 12 p>>>> \
> assume the parity bit is always for the horizontal row it's associated> with.>>>> \
> The UNRAID system still uses the data on the other drives as if it was>> striped \
> for purposes of creating the parity disk, the difference is that> the>> data is not \
> actually striped.hope this helps.> > Interesting functionality. The striping in \
> RAID 4 will give you better> throughput since several disks can be accessed at the \
> same time. But if> the disks are fast enough, as you mentioned in a previous post, \
> then> striping won't buy you anything. The Unraid gives you the advantage of> being \
> able to use different sized disks, which is cool. An updated picture> would look \
> like this:> > UNRAID> 01 05 11 p> 02 06 12 p> 03 07 13 p> 04 08 14 p> 09 15 p> 10 \
> p> > The parity disk must therefore be at least as big as the largest data disk.> > \
> RAID 4 requires the same sized disks. If the disks aren't the same size,> it only \
> uses the amount of space on each disk equal to the smallest of the> disks. The rest \
> is wasted.> > I would prefer that this proposed functionality be referred to as \
> "Unraid"> or something other than "RAID 4 without striping". RAID 4 is a technical> \
> specification that includes striping. When I see "RAID 4" I automatically> think of \
> striping. The new RAID layout is not RAID 4. It's more like> concatenation with \
> parity.> > EVMS could add a plug-in for this RAID format, but it won't be useful> \
> without the corresponding code in the kernel. EVMS could provide a patch> to the \
> kernel, but it is best to just have the code in the kernel to begin> with. Anyone \
> up for writing the kernel code and getting it accepted in the> mainline kernel? :)> \
> > Steve D.> > > > -------------------------------------------------------> This \
> > SF.net email is sponsored by: Splunk Inc. Do you grep through log files> for \
> > problems? Stop! Download the new AJAX search engine that makes> searching your \
> > log files as easy as surfing the web. DOWNLOAD SPLUNK!> \
> > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642> \
> > _______________________________________________> Evms-devel mailing list> \
> > Evms-devel@lists.sourceforge.net> To subscribe/unsubscribe, please visit:> \
> > https://lists.sourceforge.net/lists/listinfo/evms-devel
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/


[Attachment #3 (text/html)]

<html>
<head>
<style>
P
{
margin:0px;
padding:0px
}
body
{
FONT-SIZE: 10pt;
FONT-FAMILY:Tahoma
}
</style>
</head>
<body><P >Yeah, i also think this would be very good functionality to add to the \
Linux <FONT class="" >md</FONT> driver (and subsequently to EVMS).</P> <P >&nbsp;</P>
<P >For more information on this, there's a pretty huge thread over at <FONT class="" \
>avsforum</FONT> about this: <A \
> href="http://www.avsforum.com/avs-vb/showthread.php?t=573986" \
> >http://www.avsforum.com/avs-vb/showthread.php?<FONT \
> > class="">t</FONT>=573986</A>.&nbsp; One of the main guys running the company that \
> > markets this (<FONT class="" >limetech</FONT>) has posted a lot to that \
> > thread.&nbsp; I posted to it a while back asking if there were any plans <FONT \
> > class="" >GPL</FONT> the driver code, and he's said that it's something they're \
> > planning to do that (releasing the driver code, but not the management \
> > interface).&nbsp; See the following post in the thread: <A \
> > href="http://www.avsforum.com/avs-vb/showthread.php?p=6500421#post6500421" \
> > >http://www.avsforum.com/avs-vb/showthread.php?p=6500421#post6500421</A>.</P>
<P >&nbsp;</P>
<P >I don't know what's come of the plans to release this code, but he seems like a \
helpful chap, so it might be worth getting in contact with him.&nbsp; I don't know if \
Neil Brown would be interested in adding this functionality to the kernel?<BR ><BR \
></P> <P>
<HR id=stopSpelling>
</P>
<P>&gt; Subject: Re: [Evms-devel] Feature request: RAID4 without stripping (only \
parity)<BR>&gt; To: cartman@medsite.info<BR>&gt; CC: \
evms-devel@lists.sourceforge.net<BR>&gt; From: steved@us.ibm.com<BR>&gt; Date: Wed, \
15 Feb 2006 09:34:50 -0600<BR>&gt; <BR>&gt; Konstantinos Skarlatos wrote on \
02/15/2006 07:18:24<BR>&gt; AM:<BR>&gt; <BR>&gt;&gt; -----BEGIN PGP SIGNED \
MESSAGE-----<BR>&gt;&gt; Hash: SHA1<BR>&gt;&gt;<BR>&gt;&gt; Unraid is different from \
RAID4, because it does no striping. It is just<BR>&gt;&gt; a parity drive, that is \
added in a system of indepependent disks of<BR>&gt;&gt; varying capacity and file \
systems. The data on the data drives are not<BR>&gt;&gt; touched by the whole \
operation, you could add a full drive on the system<BR>&gt;&gt; and the only thing \
that would happen is the recalculation of parity on<BR>&gt;&gt; the parity drive. I \
think Talik described it well:<BR>&gt;&gt; [quote]<BR>&gt;&gt; \
Traditional<BR>&gt;&gt; RAID 4<BR>&gt;&gt; 01 02 03 p<BR>&gt;&gt; 04 05 06 \
p<BR>&gt;&gt; 07 08 09 p<BR>&gt;&gt; 10 11 12 p<BR>&gt;&gt;<BR>&gt;&gt; \
Traditional<BR>&gt;&gt; RAID 5<BR>&gt;&gt; 01 02 03 p<BR>&gt;&gt; 04 05 p \
06<BR>&gt;&gt; 07 p 08 09<BR>&gt;&gt; p 10 11 12<BR>&gt;&gt;<BR>&gt;&gt; \
UNRAID<BR>&gt;&gt; 01 05 09 p<BR>&gt;&gt; 02 06 10 p<BR>&gt;&gt; 03 07 11 \
p<BR>&gt;&gt; 04 08 12 p<BR>&gt;&gt;<BR>&gt;&gt; assume the parity bit is always for \
the horizontal row it's associated<BR>&gt; with.<BR>&gt;&gt;<BR>&gt;&gt; The UNRAID \
system still uses the data on the other drives as if it was<BR>&gt;&gt; striped for \
purposes of creating the parity disk, the difference is that<BR>&gt; the<BR>&gt;&gt; \
data is not actually striped.hope this helps.<BR>&gt; <BR>&gt; Interesting \
functionality. The striping in RAID 4 will give you better<BR>&gt; throughput since \
several disks can be accessed at the same time. But if<BR>&gt; the disks are fast \
enough, as you mentioned in a previous post, then<BR>&gt; striping won't buy you \
anything. The Unraid gives you the advantage of<BR>&gt; being able to use different \
sized disks, which is cool. An updated picture<BR>&gt; would look like this:<BR>&gt; \
<BR>&gt; UNRAID<BR>&gt; 01 05 11 p<BR>&gt; 02 06 12 p<BR>&gt; 03 07 13 p<BR>&gt; 04 \
08 14 p<BR>&gt; 09 15 p<BR>&gt; 10 p<BR>&gt; <BR>&gt; The parity disk must therefore \
be at least as big as the largest data disk.<BR>&gt; <BR>&gt; RAID 4 requires the \
same sized disks. If the disks aren't the same size,<BR>&gt; it only uses the amount \
of space on each disk equal to the smallest of the<BR>&gt; disks. The rest is \
wasted.<BR>&gt; <BR>&gt; I would prefer that this proposed functionality be referred \
to as "Unraid"<BR>&gt; or something other than "RAID 4 without striping". RAID 4 is a \
technical<BR>&gt; specification that includes striping. When I see "RAID 4" I \
automatically<BR>&gt; think of striping. The new RAID layout is not RAID 4. It's more \
like<BR>&gt; concatenation with parity.<BR>&gt; <BR>&gt; EVMS could add a plug-in for \
this RAID format, but it won't be useful<BR>&gt; without the corresponding code in \
the kernel. EVMS could provide a patch<BR>&gt; to the kernel, but it is best to just \
have the code in the kernel to begin<BR>&gt; with. Anyone up for writing the kernel \
code and getting it accepted in the<BR>&gt; mainline kernel? :)<BR>&gt; <BR>&gt; \
Steve D.<BR>&gt; <BR>&gt; <BR>&gt; <BR>&gt; \
-------------------------------------------------------<BR>&gt; This SF.net email is \
sponsored by: Splunk Inc. Do you grep through log files<BR>&gt; for problems? Stop! \
Download the new AJAX search engine that makes<BR>&gt; searching your log files as \
easy as surfing the web. DOWNLOAD SPLUNK!<BR>&gt; \
http://sel.as-us.falkag.net/sel?cmd=lnk&amp;kid=103432&amp;bid=230486&amp;dat=121642<BR>&gt; \
_______________________________________________<BR>&gt; Evms-devel mailing \
list<BR>&gt; Evms-devel@lists.sourceforge.net<BR>&gt; To subscribe/unsubscribe, \
please visit:<BR>&gt; \
https://lists.sourceforge.net/lists/listinfo/evms-devel<BR></P><br /><hr />Express \
yourself instantly with MSN Messenger! <a \
href='http://clk.atdmt.com/AVE/go/onm00200471ave/direct/01/' target='_new'>MSN \
Messenger</a></body> </html>


-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642
_______________________________________________
Evms-devel mailing list
Evms-devel@lists.sourceforge.net
To subscribe/unsubscribe, please visit:
https://lists.sourceforge.net/lists/listinfo/evms-devel

[prev in list] [next in list] [prev in thread] [next in thread] 

Configure | About | News | Add a list | Sponsored by KoreLogic