Hello Community,
I'm trying to implement a dynamic scheduler in GridSim. My idea is that
before submitting a new job, the scheduler queries every resource for
discovering their loads and then decide which one is the more suitable for
hosting new gridlets. By resource load I mean load of its machines, in a
similar sense to the contents of /proc/loadavg in any Linux system, where
the load means the average number of processes waiting to run.
I've run through examples 1 to 10 in the Gridsim. It appears that the right
way to do what it want is sending a message to the Resource with the tag
GridSimTags.RESOURCE_DYNAMICS, and store the result in a accumulator. Here
is the code that I'm running:
int id = ((Integer) super.getGridResourceList().get(0)).intValue();
super.send(id, GridSimTags.SCHEDULE_NOW, GridSimTags.RESOURCE_DYNAMICS, ID);
Accumulator accumulator = (Accumulator) super.receiveEventObject();
System.out.println("The System load after job submission: " +
accumulator.getMean());
Am I really retrieving the resource load? How is this load calculated? Does
it has any similar meaning with the values in /proc/loadavg? I'm planning to
encapsulate such behavior in a GIS, but I want to be sure that it is what I
want.
--
__________________________________
João Marcelo Uchôa de Alencar
Computer Science BSc.
jmarcelo.alencar(at)gmail.com
http://simetriade.blogspot.com
Linux User 398939
__________________________________
|
