On 4 October 2013 18:04, Stuart Sears <stuart@???> wrote:
> On 04/10/13 17:21, Daniel Llewellyn wrote:
>
> > so it's effectively, if I understand this right. not redundant AT ALL
> > in terms of a disc dying and taking the data with it.
> Yes, it is. I think you don't understand it correctly :)
>
> > AFAICT it saves two (or more) copies of each block, but it doesn't
> > ensure that those copies are on separate discs. so there could be 2
> > mirrors of a block on the same drive thereby completely negating any
> > benefit for that particular block.
>
> I don't see where you get this idea? AFAICT pains are taken to ensure
> that duplicate chunks are written to separate devices. Otherwise, what
> exactly is the point at all?
>
> man md
>
> has more information on the layouts used, in its RAID 10 section.
> Nothing I see there implies that duplicate copies of a chunk can be
> written to the same device.
>
> > It also won't improve speed of read for that particular block because
> > if the drive is busy reading another block the system can't go to
> > "another disc" to get this block because it just ain't there.
> um, yes it is.
>
> > Likewise it will halve the write speed for this particular block
> > because it can't parallelise a write to each disc independently but
> > must queue the two copies for writing to the one disc sequentially.
> Once again, no.
>
> > In short, it's a completely flawed idea IMHO (of the Not So Humble
> > variety :-p)
>
> no, not really, it's just "different".
> Although I don't use it myself.
I'm glad to be corrected :-)
--
Daniel Llewellyn
--
Please post to: Hampshire@???
Web Interface:
https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL:
http://www.hantslug.org.uk
--------------------------------------------------------------
