An executor can only be working on one thread at a time, so the manager sends threads to the idle executors, based upon the order each one becomes idle and not based upon the load of the machine running the executor.
As to actually modifying the selection of which executors to use for a task, and, say, abandoning tasks sent to 'slow' executors, that's something that hasn't been touched yet as far as I know.
Jonathan
On 22 Mar 2006 03:48:42 -0000, andrew hudson <[EMAIL PROTECTED]> wrote:
hello
will anyone tell me that how the resource discovery part in the alchemi is being implemented. I want to say that how the manager assign tasks to the executors, Is while assigning the tasks the manager check various QoS factors like latency, resource usage, bandwidth,.......etc
or it just sending threads to all the executors without considering how much load is being already there on the executor
![]()
