Hello,
I was curious about the kind of accuracy that could be expected from
the flow based model, so I decided to run an example that would be
simple enough to be computed by hand, just to see.
I simplified one of the provided examples to get a fully symmetric
backbone, with 2 flows of the same size starting at the same time.
The expected outcome is that the flows terminate at the same time,
about twice the time to send a single message without contention on
the backbone link. But instead, only one of the flows seem to share
the bandwidth on the bottleneck link while the other doesn't.
That is to say that sending a single flow takes 83.2 seconds while
when sending 2 flows, one of them takes 83.2 seconds while the other
takes 163.8.
My question here is to know whether I did something wrong in my test
case (attached to this mail), or whether this, as it seems, a bug in
GridSim. I'm using GridSim v5.2.
If someone has a fix or a workaround to solve my problem, please help.
Thanks in advance,
Mt. |