At 12:33 +0100 7/29/04, Dan Stowell wrote:
>Hi,
>
>Just about to get started with a little IP-address-based (as opposed to Rendezvous-based) Xgrid cluster. I'm going to use a G4 in my office as the controller. Is there any reason I should avoid activating the xgrid "agent" process on that same computer?
>
>I presume that the controller's duties are unlikely to make too much demand on the processor, so it would seem sensible to allow the controller to spend some of its spare time as an agent....
>
>Thanks
>Dan
With about ~120 agents, my controller tends to crash more often. So I do not use that computer as a client, or as an agent, anymore. I know not using the client made a difference. I did not test if not using the agent helped (adding 800 MHz to the 80 GHz does not make a big difference ;-). My guess is the agent should not do too much harm as it is a nice=20 process, stepping down to let other processes use the cpu (and it works really well). I don't know about potential communication bottlenecks, though.
For small-scale clusters, I would recommend using the controller as agent, which I was doing in the beginning.
charles
--
Charles PARNOT
email@hidden
Help science go fast forward:
http://cmgm.stanford.edu/~cparnot/xgrid-stanford/
Room B157 in Beckman Center
279, Campus Drive
Stanford University
Stanford, CA 94305 (USA)
Tel +1 650 725 7754
Fax +1 650 725 8021
_______________________________________________
xgrid-users mailing list | email@hidden
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/xgrid-users
Do not post admin requests to the list. They will be ignored.
