Re: [GridSim-developers] [GridSim-users] Strange error

[Giancarlo <wro...@gm...>]

Hi Marcos,
Thank you very much for your answer.
> The problem is that each entity in GridSim is a thread. Simjava2 works 
> with two event queues, namely the future event queue and the deferred 
> event queue. Every time Sim_system class reads the queue of future 
> events (i.e. at each clock tick), it puts all the next events to be 
> handles in the deferred queue. If there are multiple events scheduled 
> for the very same simulation time, they will all be placed in the 
> deferred queue. After that, Sim_system wakes up all the threads to 
> consume events.
I take a look at the code of simjava and I have another little question 
about the mechanism you have described, perhaps it's rather a confirmation
Running entities will not block on sim_get_next until they have an event 
waiting for them in the deferred queue. But an event scheduled with 
delay 0.0 by an entity during a tick run will be put in the future 
event. So the destination entity for the above event will block on 
sim_get_next and will see the event after next run of 
Sim_system.run_tick() method. Is this correct?

> Yes, this is a bug and you are not the first one to experience this 
> problem. I think people had not noticed it before because they were 
> using Java 1.4. One solution to this problem would consist in making 
> GridResource send an event to the policy instead of calling a method 
> from it. By the way, this is the approach suggested in SimJava2's 
> documentation. The other solution, if you are using the the source 
> code of GridSim instead of the jar files, is to add some 
> synchronisation to the allocation policy.
I already know about the bug, I was just wondering why the problem 
wasn't arising if the send with a port was used.
Thanks again,
Regards,
Giancarlo