Boot from Raid

[ft@angels.att.com]

On Wed, 25 Jun 1997, Margarita Suarez wrote:
> > > 2) why the 51927b block size?  Is there something special about this
> > > size?
>
> we've found that using bs=5120k means we can copy a 2gig disk in under
> 10 minutes.  51k seems awfully small...

Actually, using a block size larger than your setting of the 'maxphys'
kernel tunable, or the block size supported by the disk drives themselves
(klustsize, usually 64-256k), whichever comes first, is probably not going
to get you much, as the reads/writes will get fragmented down in the
vfs/vnode interface anyway.

Try timing this yourself by timing the transfer of a fixed amount of data
from (or to) /dev/rdsk/<somedisk> using dd and varying bs,count (assuming
your machine is fast enough that dd doesn't saturate a CPU).  You can
pretty much see where klustsize is for any given disk this way. For a
SSA112 2.1G disk and a 10MB read:

bs (k)	   time (s)
2           8.47
4           4.85
8           3.04
16          2.13
32          1.69
64          1.57
128         1.57
256         1.60
512         1.59
1024        1.56


Fred

--
Fred True				"My name is Ozymandias, King of Kings:
AT&T World Class Database Development	 Look on my works, ye Mighty,
ft@angels.att.com			 and despair!"
						-P. B. Shelley