An In-depth Benchmarking of Evolutionary and Swarm Intelligence Algorithms for Autoscaling Parameter Sweep Applications on Public Clouds
Table 8
Average makespan, average monetary cost, average number of task failures, and average L2-norm obtained by each variant of the autoscaler MOEA for applications FPA and WRF. For all metrics, lower values are better values. For each application and size, bold values are better than those obtained by MOEA-NSGA-III.
Application
Size
Autoscaler
Makespan
Cost
Task failures
L2-norm
FPA
30
MOEA-NSGA-III
6925.12
0.43
0.00
0.06
MOEA-SMPSO
9742.57
0.97
0.13
0.52
MOEA-SMS-EMOA
6869.20
2.06
0.00
0.44
MOEA-E-NSGA-III
6862.83
0.40
0.00
0.04
100
MOEA-NSGA-III
10406.34
2.13
0.00
0.17
MOEA-SMPSO
17386.56
3.38
0.53
0.78
MOEA-SMS-EMOA
10153.26
3.35
0.00
0.29
MOEA-E-NSGA-III
10221.36
1.79
0.00
0.13
300
MOEA-NSGA-III
12483.92
6.98
0.00
0.24
MOEA-SMPSO
16138.69
10.03
1.60
0.62
MOEA-SMS-EMOA
16684.92
13.50
0.00
0.71
MOEA-E-NSGA-III
11972.73
6.34
0.00
0.19
WRF
30
MOEA-NSGA-III
60741.41
11.33
0.00
0.93
MOEA-SMPSO
61388.53
9.98
3.07
0.96
MOEA-SMS-EMOA
62653.31
6.79
0.00
0.92
MOEA-E-NSGA-III
56088.10
10.02
0.00
0.81
100
MOEA-NSGA-III
60180.73
47.02
0.40
0.86
MOEA-SMPSO
63291.46
55.04
4.57
0.97
MOEA-SMS-EMOA
63530.41
24.51
3.33
0.88
MOEA-E-NSGA-III
58293.80
28.53
0.00
0.76
300
MOEA-NSGA-III
59934.56
165.42
7.07
0.92
MOEA-SMPSO
60950.30
250.11
13.37
1.06
MOEA-SMS-EMOA
64701.56
211.23
0.00
1.04
MOEA-E-NSGA-III
57568.73
111.90
0.00
0.79
indicate the best value. The baseline is in italics.
