[Bioclusters] About RSH/SSH in Cluster

Alan Ng alan at aldrich.com.my
Wed Mar 2 20:34:33 EST 2005

>>If I use ssh with key authorization,  could I need to use an empty-password 
private key ?

Yes, we do it all the time. And to make the "small" overhead smaller, you can consider using a 512-bit encryption, which is the minimum.

>>Which is better? Which is more efficient ? 
Theoretically, rsh would give better performance since it does not encrypt the communication path as everything is sent in plain-text. However, rsh is less stable than ssh (at least from my experience, thus far). Passwordless login will work for a while and then out of the blue, it'll just stop working. Happen to one of our cluster for a VR system not long ago and it had to happen during the training when the big boss of the client was around. We had to use some stupid "network problem (microsoft virus)" excuse to buy time to solve the problem. In the end, we just replace the rsh command with a symbolic link to the ssh command (had to use rsh cause the VR software insists on using it). 

Hope this helps!

"Leadership & Life-long Learning"

Alan Ng
Software Development Engineer
Linux Certified Professional
Open Source Systems Sdn Bhd
www.aldrich.com.my  Tel: 6 03-8656 0139   Fax: 6 03-8656 0132 


"Carpe Diem: Seize the Day!!"

Yun He wrote:

>Many applications in cluster need password-free environment, such as OpenPBS 
>or Torque, usually, there are two ways to let us login into different nodes 
>of cluster without giving a password, one is to use rsh with .rhost file in 
>each home directory, another is to use ssh with RSA key authorization.
>Which is better? Which is more efficient ? 
>If I use ssh with key authorization,  could I need to use an empty-password 
>private key ?
>                  Jarod
>Bioclusters maillist  -  Bioclusters at bioinformatics.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: alan.vcf
Type: text/x-vcard
Size: 202 bytes
Desc: not available
Url : http://bioinformatics.org/pipermail/bioclusters/attachments/20050303/6b5f739b/alan.vcf

More information about the Bioclusters mailing list