Re: raid10 vs raid01 type in dmraid

On Wed, Jun 27, 2007 at 10:46:49AM -0700, Gaston, Jason D wrote:
> >-----Original Message-----
> >From: Heinz Mauelshagen [mailto:[email protected]] 
> >Sent: Tuesday, June 26, 2007 12:58 AM
> >To: Gaston, Jason D
> >Cc: [email protected]; ATARAID (eg, Promise Fasttrak, 
> >Highpoint 370) related discussions
> >Subject: Re: raid10 vs raid01 type in dmraid
> >
> >On Mon, Jun 25, 2007 at 10:18:14AM -0700, Gaston, Jason D wrote:
> >> >
> >> >Eg, PC Magazine:
> >> >"RAID 10, RAID 100 - Speed and Fault Tolerance
> >> >RAID 10 is RAID 1 + 0. The drives are striped for performance 
> >> >(RAID 0), and all striped drives are duplicated (RAID 1) for 
> >> >fault tolerance."
> >> >
> >> >Yes, I konw, there's different (aargh!) definitions out there.
> >> >
> >> >Like I said: the top -> bottom of the stack definition is preferable
> >> >to me (you reach the mirror first and the stripe second 
> >while walking
> >> >the stack from the top to the bottom).
> >> >
> >> >We just got to hold on to one definition in dmraid,
> >> >
> >> >Heinz
> >> >
> >> 
> >> Ok, so the dmraid definition is not going to change.
> >
> >I'ld much rather avoid it :-)
> >
> >> 
> >> Do you see an advantage in putting the RAID1 vs. RAID0 on 
> >the bottom for
> >> being able to perform rebuilds easier?  I need to decide 
> >which way to do
> >> with isw.
> >
> >Talking about either a mirror on top of stripes or a stripe on 
> >top of mirrors,
> >it is merely a question of IO minimization on rebuilds.
> >
> >Ie. when a stripe below a mirror breaks, you've got to 
> >resilver the whole
> >stripe set, hence writing 2 or more drives rather than just one disk in
> >a mirror set beneath a stripe.
> >
> >
> >With respect to ease of rebuilds at the programming level, there's no
> >difference in changing the mapping of a top level mirror to do a full
> >resync vs. a bottom level mirror.
> >
> >
> >Getting back to your decision: you could support both ;-)
> >
> >Cheers,
> >Heinz
> >
> >> 
> >> Thanks,
> >> 
> >> Jason
> >
> 
> After adding support for RAID10 to isw.c, what will we need to do in
> order for this to be supported in anaconda, so that we can install to a
> RAID10 nested volume?  I don't see support for nested RAID levels in
> dmraid.py.

This is merely a mkinitrd issue:
we want to change mkinitrd to make initrd call dmraid directly rather
than using static tables for multipath, mirroring, dmraid, ....

With that solution, RAID10 *should* just work with anaconda.

Please talk to Peter Jones ([email protected]) about details.

> 
> Thanks,
> 
> Jason

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Red Hat GmbH
Consulting Development Engineer                   Am Sonnenhang 11
Storage Development                               56242 Marienrachdorf
                                                  Germany
[email protected]                            PHONE +49  171 7803392
                                                  FAX   +49 2626 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-