Advances, Systems and Applications
Table 4 Sampling time and training time of EdgeTuner and real Kubernetes cluster under different workloads
From: Fast DRL-based scheduler configuration tuning for reducing tail latency in edge-cloud jobs
Scenario | Sampling time (seconds) | Training time (seconds) | ||
|---|---|---|---|---|
| Â | EdgeTuner | Kubernetes | EdgeTuner | Kubernetes |
1 | 110 | 332160 | 6224 | 742708 |
2 | 118 | 342578 | 7002 | 774568 |
3 | 112 | 351027 | 7016 | 780246 |
4 | 155 | 341463 | 8189 | 896439 |
5 | 164 | 349873 | 8964 | 914782 |
6 | 159 | 356047 | 8902 | 913026 |
7 | 583 | 673240 | 9336 | 882708 |
8 | 595 | 688452 | 10402 | 943258 |
9 | 585 | 680064 | 9408 | 866802 |
10 | 587 | 682247 | 9386 | 856349 |
11 | 596 | 684065 | 9502 | 876543 |
12 | 740 | 1114389 | 16134 | 1446439 |
13 | 780 | 1156000 | 17225 | 1456020 |
14 | 764 | 1175438 | 16548 | 1470865 |
15 | 796 | 1118634 | 16208 | 1450862 |
16 | 802 | 1116065 | 17106 | 1465684 |