Rounding problem in VMSchedulerTimeShared

[GoogleCodeExporter]

1.      First build up a VM and a Cloudlet, the running time of a single 
Cloudlet on a VM is one second.
2.      Create 10 Cloudlet into the VM, the running time is 10 seconds.
3.      Create 100 Cloudlet into the VM, the running time is 99.5 seconds.
4.      Create 1000 Cloudlet into the VM, the running time is 915 seconds.

The above experiments are all conducted using the TimeShared mode.

My question is: When there was one VM, my understanding towards
TimeShared is to distribute time period, to distribute the running
time to a single person. Take 1000 missions for example, is to divide
the time unit into 1000 periods, let each Cloudlet take 1 period VM
resource, if so, 1000 single missions running separately each require
1 second, their total running time shouldn't be less than 1000
seconds, let alone the consumption of communication between task

Original issue reported on code.google.com by rodrigo.calheiros on 31 Oct 2011 at 11:31

[GoogleCodeExporter]

The issue was caused by rounding problems. To fix it, ResCloudlet now processes 
Instructions rather than MI.

Original comment by rodrigo.calheiros on 15 Dec 2011 at 5:42

• Changed state: Fixed