George,
I quote Legato PSE
" He said that, from a performance perspective, it's best to make sure that the
fragments (head, middle section or tail) are on
contiguous media files on the volume. Tape rewind and positioning is expensive
and detrimental for performance and should be
avoided during the read part of the clone. Needless to say if all the fragments
are contiguous on the same volume (no spanning) the
search for the segments and clone would be the fastest. "
What I do is clone complete saveset's 1st by volume (mminfo -omo ) with the
sumflags = c , then I clone any saveset that spans
volumes.
mminfo -t${fromdate} -omo -q"$minusq" -r'ssid,sumflags' | tail +2 | \
while read ssid sumflag; do
sumflag1=$(echo $sumflag | cut -c1)
case ${sumflag1} in
"c")
# The complete saveset is on this volume, so do it 1st
echo $ssid >> $tmpssidonly
;;
"h")
# Only the header is on this volume
echo $ssid >> $tmpssidonly_span
;;
*)
;;
esac
done
-----Original Message-----
From: George Sinclair [mailto:George.Sinclair AT NOAA DOT GOV]
Sent: Friday, September 26, 2003 1:39 PM
Subject: Re: Faster Cloning. By saveset or volume?
I seem to recall that this got into this issue of de-multiplexing versus
maintaining the multi-plexing. Then again, that may be have
been the difference between running the nsrclone command with a list of saveset
ids (on the command line with -S or via an input
file as -S -f file) versus iterating over the list one at a time. I think if
you iterate it, the savesets would have to be
un-multiplexed, but if you don't iterate it one at a time then the
multi-plexing is preserved. If that's true, then it seems to me
that cloning the entire volume would have to be faster because it would not
have to de-multiplex anything, but if you did it by
saveset, one at a time, then it would, unless you ran it using an input file
with all the savesets in there.
So, my question would be: is running nsrclone with an input file that contains
all the savesets tantamount to cloning the whole
volume in terms of over all time. In other words, which of the following is
faster:
nsrclone -s server -b pool volname
nsrclone -s server -b pool -S -f file
where file contains all the saveset ids on the volume.
I mean, both operations should amount to the same thing, but I'm thinking that
cloning the volume using the first command is easier
and could not be slower than the second.
George
"Novello, Guy" wrote:
>
> Hello Everyone,
>
> I seem to remember a discussion over which was the faster way to
> clone. I am currently using the ssid, but I am thinking it may be
> faster cloning the volumes.
>
> Has anyone tested this?
>
> Thanks in Advance!
>
> Guy
>
> --
> Note: To sign off this list, send a "signoff networker" command via
> email to listserv AT listmail.temple DOT edu or visit the list's Web site at
> http://listmail.temple.edu/archives/networker.html where you can also
> view and post messages to the list.
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
--
Note: To sign off this list, send a "signoff networker" command via email to
listserv AT listmail.temple DOT edu or visit the list's Web
site at http://listmail.temple.edu/archives/networker.html where you can also
view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|