[GE users] Help configuring grid to use ssh instead of rsh

Magnus Söderberg magnus.soderberg at switchcore.com
Fri Apr 8 09:32:43 BST 2005


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Gary Thomas wrote:
> The jobs generate heavy network traffic because they are typically
> compiling sources off of a network file server.  Each machine has 1
> network card.
Running Linux? Perhaps Redhat WS3.0 or similar?
We have noted an issue with WS3.0 such that if we run 2 compile jobs on a dual-cpu 
machine, it seems that there are no CPU-cycles left for networking/NFS.
This was less bad with RH7.3. We have filed an issue with RedHat, but to describe the 
response as fast would be, hmmm let me see, "unfair" perhaps. Well, you get what you pay for.

Anyhow, we have noted that it is a lot more efficient to first copy the files to a local 
disk, compile and then copy the result back.

Another thing that slows things is having LD_LIBRARY_PATH set unnecessarily. SGE sets it 
itself, as have been discussed previously on this list, but not having anything 
unnecessary there might also help things, perhaps clearing/reducing it at the beginning of 
the job-script. Also check out the HOWTO on reducing unnecessary NFS traffic, by keeping 
at least spool directories local.

I think any of these measures will help you more than switching to ssh.

regards

Magnus Söderberg, M.Sc.E.E.
Staff Engineer
SwitchCore AB        Phone +46 46 2702560
Emdalavägen 18       Fax   +46 46 2702581
SE-223 69 Lund, Sweden

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list