In this experiment, we create three child containers C1, C2 and C3 for the resource container to which init is bound, having a fixed CPU share of 16%, 32% and 48% respectively. We also create three CPU bound jobs R1, R2 and R3. The cumulative execution time of these jobs are shown in figure 5.
The figure shows that when all the processes are running, R1, R2 and R3 get 14.3%, 28.6% and 42.9% of the CPU respectively. There is a proportional decrease in the CPU allotted than what has been specified and may be due to the lower resolution of the CPU timer6, pre-emption being done only at safe points and the non-accounting of some of the activities in the system (eg. network activities that are charged to rc_system, the time taken for scheduling etc.). Once R3 exits, R1 and R2 get 25.2% and 50.1% of the CPU respectively and the remaining CPU power is used by various system activities and other processes in the system. This is more than what has been specified as there are not many other activities in the system. Once R2 also exits, R1 gets 50.0% of the CPU.
