|
Hi Alois,
Sorry for the late reply but our mailserver's disk crashed very badly.
This always happens when I start replying to mailinglists after weeks.
> But this is not the rcp I mean. (or have you been joking anyway?)
:) Yes, I'm sorry. But the answer ist still yes. Everything based
on udp or tcp should be possible to load balance. The load balancer
is based on L4 technology.
> the rcp (remote copy) I mean is one of the 'r*' services based
> on sun-rcp and portmap, as far as I know.
Yep.
> I do use rsync all the time to adminster my curent remote server, and
> will of course use it once the new LVS cluster becomes the remote server.
ok.
> But rsync uses either rsh or ssh as a 'transport protocol', and our
> rsync in the local machines is configured to use ssh, so we are
> back to the original question.
No, not really. rsync can be told to check if the the data is already
there. If so, it doesn't sync. One of our webtree has 800MB. We sync
it with rsync:
rsync --archive --update --hard-links --delete-after --force -e \"ssh -p
$sshport\""
This works pretty fast and accurate.
> ssh is slow in the local network:
> for a 9.1 Mb gzipped file I measure 7 seconds.
Hmm, what is your hardware used (the two nodes, the NICs, the kernel
and the switch)?
> rcp takes less than 1 second.
But then you have a very slow processor. Back here scp of 100MB takes
31.4 secs with 2048bits and 80% cpuload.
> For transferring 10 Gb of data, this is about 7000 seconds (2 hours)
> versus 1000 seconds (18 minutes).
Hmm, I haven't yet studied your network topology and your desired
setup but maybe we will be able to use NFS, don't know yet.
> I am reluctant to play around with key length for ssh, this might cost me
> more time than 7000 seconds...
It shouldn't take more then 70 seconds but this is debatable. I still
think we could go for rsync.
Best regards,
Roberto Nibali, ratz
--
mailto: `echo NrOatSz@xxxxxxxxx | sed 's/[NOSPAM]//g'`
|
