Re: lilo/raid?

[jw schultz <jw@pegasys.ws>]

On Tue, Jul 02, 2002 at 11:12:03AM +0200, Zwane Mwaikambo wrote:
> On Tue, 2 Jul 2002, Helge Hafting wrote:
> 
> > > > /dev/md1                swap                    swap    defaults        0 0
> > > 
> > > One small thing, you do know that you can interleave swap?
> > 
> > There are sometimes reasons not to do that.
> > Heavy swapping may be caused by attempts to cache 
> > massive io on some fs.  You better not have swap
> > on that heavily accessed spindle - because then
> > everything ends up waiting on that io.
> > 
> > Keeping swap somewhere else means other programs
> > just wait a little for swap - undisturbed by the massive
> > io also going on.
> 
> True, but what i meant was that instead of creating a RAID device to swap 
> to, he could have just interleaved normal swap partitions and gotten the 
> same effect.

Not a config i would recommend.  While spreading swap over
multiple spindles that way is good for speed it multiplies the
likelihood that a disk failure will down the system.  If you
think a dead filesystem is a mess, just watch what happens
when swap goes dead.

Much better to put swap on several RAID-1 volumes if
you want to spread it around.  Disks just aren't that
expensive anymore.  RAID-5 though would be a bad move since
swap gets an order of magnitude more writes than reads.  On
account of the heavy write tendencies i lean toward
so-called hardware RAID for swap in order to offload the PCI
buss.

-- 
________________________________________________________________
	J.W. Schultz            Pegasystems Technologies
	email address:		jw@pegasys.ws

		Remember Cernan and Schmitt