Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 31 Jan 1998 15:55:27 -0600
From:      Karl Denninger  <karl@mcs.net>
To:        Brian Tao <taob@nbc.netcom.ca>
Cc:        current@FreeBSD.ORG
Subject:   Re: RAID controllers - folks, check this thing out
Message-ID:  <19980131155527.19192@mcs.net>
In-Reply-To: <Pine.GSO.3.95.980131161942.27817Z-100000@tor-adm1>; from Brian Tao on Sat, Jan 31, 1998 at 04:28:00PM -0500
References:  <19980131144604.03410@mcs.net> <Pine.GSO.3.95.980131161942.27817Z-100000@tor-adm1>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jan 31, 1998 at 04:28:00PM -0500, Brian Tao wrote:
> On Sat, 31 Jan 1998, Karl Denninger wrote:
> > 
> > CCD mounted async is NOT a fair comparison - not at all.
> 
>     I note that in my message.  Async mounts delay metainfo writes,
> and the RAID controller delays all writes to disk via its cache.  It
> would be unfair to CCD, in that your chances of recovering a busy
> async filesystem after a crash are nearly zero.  The comparison is
> useful when you consider that the CMD is almost able to keep up even
> without asynchronous updates.

Yes, but it doesn't quite get there, and you lose hot-rebuild.  Also, CCD is
only RAID 0 or RAID 1 - where there is no parity computation.  The CMD
controler in a RAID 0+1 configuration might be a LOT faster.

> > Also, RAID 5 with other than 5 disks is sub-optimal, and some of those
> > tests were run with fewer drives.
> 
>     What is magic about having 5 drives?  Obviously the more drives
> you add, the better the performance.  I limited both RAID units to
> three drives to provide a more level comparison.  These were done on
> eval units, and I would have liked six or more drives.
> -- 
> Brian Tao (BT300, taob@netcom.ca)
> "Though this be madness, yet there is method in't"

RAID 5, due to the way it stripes parity across the volumes, has a "sweet
spot" in performance at 5 spindles.

It is significantly slower with either more or fewer devices.

RAID 4 keeps all parity info on one disk, and doesn't have that problme (but
it DOES have the problem of the single spindle being a bottleneck on
writes).

--
-- 
Karl Denninger (karl@MCS.Net)| MCSNet - Serving Chicagoland and Wisconsin
http://www.mcs.net/          | T1's from $600 monthly to FULL DS-3 Service
			     | NEW! K56Flex support on ALL modems
Voice: [+1 312 803-MCS1 x219]| EXCLUSIVE NEW FEATURE ON ALL PERSONAL ACCOUNTS
Fax:   [+1 312 803-4929]     | *SPAMBLOCK* Technology now included at no cost



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19980131155527.19192>