On Tue, Nov 28, 2006 at 09:34:51AM -0800, Chuck Swiger wrote:
> On Nov 27, 2006, at 6:03 PM, Aaron J. Grier wrote:
> >the whole dump -> rmt -> tape write interaction ends up going in fits
> >and starts.
> 
> Agreed.  Note that you can use dd to rebuffer the output of dump,
> which will speed things up considerably, assuming that your version
> of dump does not support configuring the blocksize it uses.

NetBSD's dump is limited to writing 64K blocks (128 512-byte "tape
blocks").

I get roughly 600kB/s to my DLT8000 through ssh/rmt with the following
invocation:

	/sbin/dump -1u -a -b 128 -h 0 -k 64 -r 128 $FILESYSTEM

dumping the root filesystem from my alpha to an i386 box (/dev/null)
with dump vs star I get the following:

dump to /dev/null: 4203 KB/s
star to /dev/null: 4116.75 KB/s
dump with BSD rmt: 960 KB/s
dump with star rmt: 989 KB/s
star with star rmt: ~1300 KB/s
star with bsd rmt server: 1388.26 KB/s

both with 64kB write blocks.  I'll have to try reblocking with dump to
see how much difference that makes; thanks for the suggestion.  I know
the DLTs can handle very large blocks, up to 16MB I think.

-- 
  Aaron J. Grier | "Not your ordinary poofy goof." | agrier@poofygoof.com
              "silly brewer, saaz are for pils!"  --  virt