Hello, I don’t use JPPF yet, but I plan to use quite soon and I have small question about it:
I have a lot of processing tasks. Each of them takes about 2-8 hours on one PC (3GHz, 1Gbt). If I send such tasks for processing to clients and clients will process it for such a long time, will JPPF work properly? Won’t I meet any timeout related problems?
Thanks in advance for your answer.
p.s.: Small suggestion. It is common, that all client computers will work at the night when office is empty, but when people back they need their computers for work. It would be nice if they can suspend processing somehow (including closing client-program) and recover it back when they go home.
Pretty good question.
I'm ashamed to admit we haven't tested this use case yet (I mean not for 8hrs straight).
The only thing that might be impacted is the socket connection between node and server. If your network doesn't break somehow during the night, then the connection should be maintained.
If it does break, then the task will be resubmitted and whatever work had been done on this task will be lost.
If possible at all, I would recommend splitting it in much smaller tasks.
Regarding the capacity to stop the tasks when people come back to work, I believe the screensaver version of the node is a perfect fit. You can download it in the files section (http://sourceforge.net/project/showfiles.php?group_id=135654).
We currently only have a Windows version, and we're still working hard on making the Linux version work as well.
I hope this answers your questions, let me know if there anything else you need,
Well, I wanted to add another thing: I have actually tested the framework for long periods of time (8 to 14hrs). I'm using genetic algorithms to evolve populations of neuroal networks, and that can generate some rather large computations.
However, what I haven't done, is to test with an inactive connection for that amount of time, since I rather had many small tasks that took a few seconds to execute.