Power-Aware EDZL Scheduling upon Identical Multiprocessor Platforms

On-Line Dynamic Voltage Scaling for EDZL Scheduling on Symmetric Multiprocessor Real-Time Systems

“…In previous work of [11], the method of dynamic speed calculation for EDZL was applicable only when all the active jobs can be completed before ℛ min ( ). Thus the previous method could be applied for very limited cases only.…”

“…The active job set (1) is 1 (1) = (1,5), 2 (1) = (4,12), 4 (1) = (2,8). For this job set, the dynamic speed determination method presented in [11] cannot be applied because However, if we re-assign the absolute deadlines of jobs as 1 (1) = 4, 2 (1) = 5, 4 (1) = 4 at time 1, then the job 2 (1) reaches zero laxity at time 1 and the job set is scheduled by EDZL as shown in Figure 1(b). As a result, the latest completion time of jobs in (0) becomes shorter than Figure 1 In the above example, we found that an active job set can be completed with absolute deadline re-assignment.…”

“…For EDZL scheduling, we proposed low-power algorithms for static speed determination and dynamic speed determination in [11,12]. To determine a static speed, we was used the schedulability condition of EDZL and the schedulability condition of EDF.…”

“…Firstly, we tested the performance when the tasks run with the worst case execution. We compared the energy consumption of the proposed algorithm with the algorithm proposed in [11]. Figure 2 and Figure 3 shows simulation results for G1 and G2 when the tasks run with the worst case execution.…”

“…Based on the observation that tasks might complete earlier than their worst-case execution time, a slack time reclamation algorithm was proposed in [5]. Though the global EDZL [6] outperforms EDF [7][8][9][10], few works on energy conservation for EDZL other than our previous works in [11,12] could be found. In previous works, we proposed an off-line algorithm and an on-line algorithm for EDZL; the off-line algorithm determines a static processor speed at design time and the on-line algorithm determines a dynamic speed at each scheduling point.…”