[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ssh-vs-rsh benchmark result

Rewinding to the original post, I have a fundamental question:

Junichi Uekawa wrote:


I've tried running ssh and rsh on a local cluster, for your interest.

The experiment was ran on 60 nodes of a idle PC cluster,
each node is a dual-Pentium III 1GHz, memory 512MB,
interconnected with 100BASE-TX connection.

I have timed the following cases:

dsh -a w 3.748 -- running "rsh w" on all hosts sequentially dsh -a -c w 1.170 -- running "rsh w" on all hosts parallelly

dsh -a -rssh w   23.051 -- running "ssh w" on all hosts sequentially
dsh -a -rssh -c w  4.306 -- running "ssh w" on all hosts parallelly

So clearly, ssh connections take longer to start.

But more importantly for us Beowulf users, who might run a job for days and not care about a few seconds of startup time, I have been under the impression that ssh encrypts all communications during the session, which would use considerable CPU time in a parallel job with lots of communication. The more communication and faster the connections (e.g. gigabit), the more CPU time would have to be devoted to this task. Based on this understanding, I've been putting my Beowulfs on private nets and using rsh.

Is my basic understanding sound, or does ssh send the session info, X stuff, etc. in the clear? Or does this not add significant computation time, e.g. on the order of what the kernel uses to get the packets out through the NIC? Has anyone seen/done relevant benchmarks which would test the magnitude of this effect?


-Adam P.

GPG fingerprint: D54D 1AEE B11C CE9B A02B  C5DD 526F 01E8 564E E4B6

Welcome to the best software in the world today cafe! <http://lyre.mit.edu/%7Epowell/The_Best_Stuff_In_The_World_Today_Cafe.ogg>

Reply to: