Progress Report 2013/08/22
Model Modification Each core works under the same frequency due to hardware limitation. A task can have different processing rates during its execution.
Observation 1 If there is only one task, the processing rate should remain unchanged. => w = 0 or 1 w 1-w
Processing rate According to observation 1, if there is only one task, the cost is: Assume that p j leads to the smallest cost. p j is independent with B.
Length of Overlapping Assume that tasks have different arrival time. Increase the processing rate of Task 0 during t 0 to t 1 reduces the overlapping of Task 0 and Task 1. Task 0 t1t1 t0t0 Task 1 Task 0 Task 1 t2’t2’t2t2
Length of Overlapping(Cont.) However, if p j is independent with B, the blue parts of Task 1 must have the same processing rate. Task 0 t1t1 t0t0 Task 1 Task 0 Task 1 t2’t2’t2t2
Queuing Theory Compute the average waiting time of tasks, and adjust the processing rate. CPU (1-p) p I/O λ μ
Concern Scheduling under “each core with the same frequency” makes our work no difference with the others.
Disccusion