sftp/scp performance testing

[mandar at webchat.chatsystems.com]

For curiosity's sake, how does sftp/scp performance compare when the
underlying protocol is changed i.e rsa1 vs rsa2 vs dsa keys for
encryption?

- Mandar


On Sun, 6 Jan 2002, Dan Kaminsky wrote:

> Date: Sun, 6 Jan 2002 22:01:21 -0800
> From: Dan Kaminsky <dan at doxpara.com>
> To: Tobias Ringstrom <tori at ringstrom.mine.nu>, mike at perham.net
> Cc: openssh-unix-dev at shitei.mindrot.org
> Subject: Re: sftp/scp performance testing
>
> > Weird.  You need to run the patched sshd, ssh, and sftp programs.
> > Running on localhost, your CPU utilization should be 100%.  You can verify
> > this using vmstat or top.
>
> A small note --
>
>     My method for testing OpenSSH performance has been something along the
> lines of:
>
>     cat /dev/zero | ssh user at host "cat > /dev/null"
>
>     and then monitoring network traffic using another tool.  Something like:
>
>     cat /dev/zero | cpipe | ssh user at host "cat > /dev/null"
>
>     would also work.  This method is useful because the zeronum generator is
> faster than any randomizer, and it completely filters out disk I/O.
>
>     Markus, a few words on OpenSSH profiling might be helpful? :-)
>
> --Dan
>
>
> _______________________________________________
> openssh-unix-dev at mindrot.org mailing list
> http://www.mindrot.org/mailman/listinfo/openssh-unix-dev
>