Date: Tue, 28 Oct 2003 14:58:29 +0300 (MSK) From: Dmitry Morozovsky <marck@rinet.ru> To: FreeBSD-gnats-submit@FreeBSD.org Cc: grog@FreeBSD.org Subject: docs/58615: update for Vinum chapter of Handbook: describe raid5 plexes Message-ID: <200310281158.h9SBwTrD066612@woozle.rinet.ru> Resent-Message-ID: <200310281200.h9SC0Zg4043149@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 58615
>Category: docs
>Synopsis: update for Vinum chapter of Handbook: describe raid5 plexes
>Confidential: no
>Severity: non-critical
>Priority: medium
>Responsible: freebsd-doc
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: doc-bug
>Submitter-Id: current-users
>Arrival-Date: Tue Oct 28 04:00:34 PST 2003
>Closed-Date:
>Last-Modified:
>Originator: Dmitry Morozovsky
>Release: FreeBSD 4-STABLE i386
>Organization:
Cronyx Plus LLC (RiNet ISP)
>Environment:
System: FreeBSD 4-STABLE
>Description:
Vinum Chapter (Number 13 as of current Handbook) mostly had been written in the
epoch of FreeBSD 3.x, where raid5 Vinum capabilities were strict-licensed. This
is no longer the case for RELENG_4; however, raid5 has been described in
"conceptual" sections but not in "implementation" ones. Here is update for two
level1 sections.
Next update for Vinum chapter should be bringing real configuration examples to
consistent state. But I suppose it should be different PR.
>How-To-Repeat:
>Fix:
Index: chapter.sgml
===================================================================
RCS file: /home/ncvs/doc/en_US.ISO8859-1/books/handbook/vinum/chapter.sgml,v
retrieving revision 1.24
diff -u -r1.24 chapter.sgml
--- chapter.sgml 31 Aug 2003 16:41:36 -0000 1.24
+++ chapter.sgml 28 Oct 2003 11:51:27 -0000
@@ -362,7 +362,7 @@
<sect2>
<title>Which Plex Organization?</title>
<para>The version of Vinum supplied with FreeBSD &rel.current; implements
- two kinds of plex:</para>
+ three kinds of plex:</para>
<itemizedlist>
<listitem>
@@ -386,11 +386,24 @@
(fractionally) more complex code and restrictions on
subdisks: they must be all the same size, and extending a
plex by adding new subdisks is so complicated that Vinum
- currently does not implement it. Vinum imposes an
+ currently does not implement it (actually, you have to reconstruct
+ file system after adding or removing subdisks to the underlying plex).
+ Vinum imposes an
additional, trivial restriction: a striped plex must have
at least two subdisks, since otherwise it is
indistinguishable from a concatenated plex.</para>
</listitem>
+
+ <listitem>
+ <para>raid5
+ plexes implement redundancy with low volume overhead for the price
+ of higher CPU load and relatively low write speed. As for striped
+ plexes, subdisks involved must be the same size, and extending
+ raid5 plex will need its complete reformatting. Minimal raid5 plex
+ must contain at least 3 subdisks; due to performance reasons, 5 or
+ more subdisks (all on different drives, of course) are
+ recommended.</para>
+ </listitem>
</itemizedlist>
<para><xref linkend="vinum-comparison"> summarizes the advantages
@@ -427,6 +440,15 @@
<entry>High performance in combination with highly concurrent
access</entry>
</row>
+
+ <row>
+ <entry>radi5</entry>
+ <entry>3</entry>
+ <entry>no</entry>
+ <entry>yes</entry>
+ <entry>High read performance in combination with redundancy such
+ as large data archives</entry>
+ </row>
</tbody>
</tgroup>
</table>
@@ -708,6 +730,30 @@
<graphic fileref="vinum/vinum-raid10-vol">
</figure>
</para>
+ </sect2>
+
+ <sect2>
+ <title>Resilience and Performance: Hard Way</title>
+
+ <para><anchor id="vinum-raid5">With sufficient hardware and CPU power,
+ especially when most of disc activities are reads, it would be useful
+ to make raid5 volume, which consumes only 1/n of disk volume for
+ resilience instead of 1/2 for mirrored volumes, and still achieves
+ great read performance. A typical configuration file might
+ be:</para>
+
+ <programlisting>
+ volume raid5
+ plex org raid5 512k
+ sd length 102480k drive a
+ sd length 102480k drive b
+ sd length 102480k drive c
+ sd length 102480k drive d
+ sd length 102480k drive e</programlisting>
+
+ <para>Note we have single plex here because raid5 provides
+ fault-tolerance itself.</para>
+
</sect2>
</sect1>
>Release-Note:
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200310281158.h9SBwTrD066612>