View | Details | Raw Unified | Return to bug 58615
Collapse All | Expand All

(-)chapter.sgml (-2 / +48 lines)
Lines 362-368 Link Here
362
    <sect2>
362
    <sect2>
363
      <title>Which Plex Organization?</title>
363
      <title>Which Plex Organization?</title>
364
      <para>The version of Vinum supplied with FreeBSD &rel.current; implements
364
      <para>The version of Vinum supplied with FreeBSD &rel.current; implements
365
	two kinds of plex:</para>
365
	three kinds of plex:</para>
366
    
366
    
367
      <itemizedlist>
367
      <itemizedlist>
368
	<listitem>
368
	<listitem>
Lines 386-396 Link Here
386
	    (fractionally) more complex code and restrictions on
386
	    (fractionally) more complex code and restrictions on
387
	    subdisks: they must be all the same size, and extending a
387
	    subdisks: they must be all the same size, and extending a
388
	    plex by adding new subdisks is so complicated that Vinum
388
	    plex by adding new subdisks is so complicated that Vinum
389
	    currently does not implement it.  Vinum imposes an
389
	    currently does not implement it (actually, you have to reconstruct
390
	    file system after adding or removing subdisks to the underlying plex).  
391
	    Vinum imposes an
390
	    additional, trivial restriction: a striped plex must have
392
	    additional, trivial restriction: a striped plex must have
391
	    at least two subdisks, since otherwise it is
393
	    at least two subdisks, since otherwise it is
392
	    indistinguishable from a concatenated plex.</para>
394
	    indistinguishable from a concatenated plex.</para>
393
	</listitem>
395
	</listitem>
396
397
	<listitem>
398
	  <para>raid5
399
	    plexes implement redundancy with low volume overhead for the price
400
	    of higher CPU load and relatively low write speed.  As for striped
401
	    plexes, subdisks involved must be the same size, and extending
402
	    raid5 plex will need its complete reformatting.  Minimal raid5 plex
403
	    must contain at least 3 subdisks; due to performance reasons, 5 or
404
	    more subdisks (all on different drives, of course) are
405
	    recommended.</para>
406
	</listitem>
394
      </itemizedlist>
407
      </itemizedlist>
395
    
408
    
396
      <para><xref linkend="vinum-comparison"> summarizes the advantages
409
      <para><xref linkend="vinum-comparison"> summarizes the advantages
Lines 427-432 Link Here
427
	      <entry>High performance in combination with highly concurrent
440
	      <entry>High performance in combination with highly concurrent
428
		access</entry>
441
		access</entry>
429
	    </row>
442
	    </row>
443
	    
444
	    <row>
445
	      <entry>radi5</entry>
446
	      <entry>3</entry>
447
	      <entry>no</entry>
448
	      <entry>yes</entry>
449
	      <entry>High read performance in combination with redundancy such
450
	        as large data archives</entry>
451
	    </row>
430
	  </tbody>
452
	  </tbody>
431
	</tgroup>
453
	</tgroup>
432
      </table>
454
      </table>
Lines 708-713 Link Here
708
	  <graphic fileref="vinum/vinum-raid10-vol">
730
	  <graphic fileref="vinum/vinum-raid10-vol">
709
        </figure>
731
        </figure>
710
      </para>
732
      </para>
733
    </sect2>
734
735
    <sect2>
736
      <title>Resilience and Performance: Hard Way</title>
737
738
      <para><anchor id="vinum-raid5">With sufficient hardware and CPU power,
739
	especially when most of disc activities are reads, it would be useful
740
	to make raid5 volume, which consumes only 1/n of disk volume for
741
	resilience instead of 1/2 for mirrored volumes, and still achieves
742
	great read performance.  A typical configuration file might
743
	be:</para>
744
745
      <programlisting>
746
	volume raid5
747
      plex org raid5 512k
748
        sd length 102480k drive a
749
        sd length 102480k drive b
750
        sd length 102480k drive c
751
        sd length 102480k drive d
752
        sd length 102480k drive e</programlisting>
753
754
      <para>Note we have single plex here because raid5 provides
755
	fault-tolerance itself.</para>
756
711
    </sect2>
757
    </sect2>
712
  </sect1>
758
  </sect1>

Return to bug 58615