Verlust - Free Memory

Submitted by jondr on Wed, 29/08/2007 - 10:23.

Sorry for the double post... too much latency and process timed out at this end.
I am using IG 1.0.
I am running IG with a server and one client. When I run the POV Ray application, I monitor with Ganglia. I notice that free memory gradually decreases. This memory is not recovered after a the computation is finished. When I start another POV Ray computation, the free memory continues to decrease and the computation becomes very slow. Eventually computation ceases, nothing happens, and I must reboot the grid.

Is it possible that there is a memory leak in the POV Ray application that causes this? I hope it is not in the Globus Toolkit or your interface programs. Have you experienced this problem? Is there anything that can be done to prevent this?

Thank you for your attention to my query.

Jon Richards, Operations Research Analyst, Fort Collins, Colorado, USA.
richardsjd@acm.org

How much memory & which POV-Ray frontend

Hello,

at first it would be good to know, how much memory your server and how much your client has.

Did you use the POV-Ray portlet in Gridsphere or did you use the php-script page?

When writing to the (virtual) disk in knoppix (like rendered pictures, temp files etc), you have to remember, that all this costs also memory, because everything runs from a ramdisk.

However, maybe you found a real problem. So please tell a bit more detailed, what happened and we try to reproduce it here.

Thanks

Tim

reply: memory and application

The server has 1.5GB and the client has only 512MB.
I have only been using the POV-Ray from the main IG drop-down menu, I suppose this is the php-script page.
I am running the third computation this morning. I see from Ganglia that the client (Bombay) has only 3m of 2.4GB disk space free. CPU-user load is less than 10%. I see the server has the same readings.
mem-Free on Client is 55KB and on the Server it is 104KB. I'm not familiar yet with running Knoppix, however I can now begin to understand the implications. Perhaps this is normal operation.
I will try going through Gridsphere and see how this goes. I'll also try to find some more RAM to give more room for operation.
thank you

Jon Richards
Operations Research Analyst
Fort Collins Science Center
U.S. Geological Survey
Fort Collins, Colorado, USA