rsync vs. rcp

[jw schultz]

On Wed, Feb 19, 2003 at 09:53:05PM -0000, va_public <yahoogroups at vikas.mailshell.com> wrote:
> --- In rsync at yahoogroups.com, Donovan Baarda <abo at m...> wrote:
> > On Thu, 2003-02-20 at 05:55, va_public  wrote:
> > 
> > RSYNC DOES NOT WORK WITH 1GB+ FILES... unless you have a 
> sufficiently
> > large block size. See the following;
> > 
> > http://www.mail-archive.com/rsync@l.../msg05219.html
> 
> OK. I read the thread. Pretty interesting design discussion on rsync 
> internals.
> 
> Has any of it been implemented in 2.5.6? When is it planned to 
> implement it?
> 
> So, bottom line, you are telling me, dont use rsync for my Oracle 
> database backups?
> 
> Thats a bummer. Oracle database files are pretty sparse and, at a 
> block level, very few things change, so I was hoping that 
> using 'rsync' instead of 'rcp' to do my daily backups would be 
> IMMENSELY faster.
> 
> As per the final messages on the thread, just increasing the csum 
> from 2 to 4 bytes seemed to miraculously solve the problem, right? So 
> why isnt this patch being included?

Dynamic block size is (at least to a limited extent) already
implemented.  Despite what the manpage says dynamic
checksums length does not appear to be implemented yet.

Setting aside ssh vs rsh issues.  Rsync is designed to
optimize for network bandwidth.  To do so it thrashes the
disks, CPU and memory.  If your network is fairly fast rsync,
even with --whole-file, will be slower than a straight copy.
I've been in situations where the network/disk speed ratio
was such that copying over the network (cpio over rsh pipe)
was faster local disk-to-disk copy.

As for oracle tablespaces, they benefit from a block size
that correlates with the oracle page-size.  There is another
recent thread that discusses oracle backups.

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

		Remember Cernan and Schmitt